<%BANNER%>

Laser Induced Breakdown Spectroscopy As a Tool for Discrimination and Quantitative Characterization of Glass for Forensi...

Permanent Link: http://ufdc.ufl.edu/UFE0024353/00001

Material Information

Title: Laser Induced Breakdown Spectroscopy As a Tool for Discrimination and Quantitative Characterization of Glass for Forensic Applications
Physical Description: 1 online resource (162 p.)
Language: english
Creator: Rodriguez Celis, Esperanza
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: Chemistry -- Dissertations, Academic -- UF
Genre: Chemistry thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Material analysis and characterization can provide important information as evidence in legal proceedings. Taking advantage of the multi-element detection capability and minimal to no sample preparation of LIBS, this technique is proposed as a viable alternative for glass analysis. Discrimination studies by linear and rank correlation methods were performed to glasses from automobile windows. Linear correlation combined with the use of a spectral mask, which eliminates some high-intensity emission lines from the major elements present in glass, provides effective identification and discrimination at a 95% confidence level. In the course of this study, several key instrumental parameters were identified and investigated, and a comparative study regarding the performance of up to four different commercial instruments for LIBS analysis of solids has been made. Results indicate that the spectral resolution and sensitivity of the systems were linked to the performance and suitability of the technique for material identification by correlation methods, especially when samples were of very similar composition. While the qualitative characterization of materials without sample preparation is certainly one of the main advantages of LIBS, the possibility of performing accurate quantitative analysis relies on the use of calibration curves made with matrix-matched standards. Such quantitative analysis would also improve the discrimination capability of the technique in the case of the glass samples analyzed. Quantitation is one of the analytical aspects of LIBS where improvements are needed. In addition to the conventional analytical procedures followed in LIBS analysis, a normalization procedure based upon the possible correlation existing between the background fluctuation and the analytical signal. This method has been applied to several types of solid samples. The results of all these investigations have allowed a fair assessment of the applicability of LIBS for discrimination and quantitative characterization of solid samples in general, and glasses in particular.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Esperanza Rodriguez Celis.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Omenetto, Nicolo.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2009-11-30

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024353:00001

Permanent Link: http://ufdc.ufl.edu/UFE0024353/00001

Material Information

Title: Laser Induced Breakdown Spectroscopy As a Tool for Discrimination and Quantitative Characterization of Glass for Forensic Applications
Physical Description: 1 online resource (162 p.)
Language: english
Creator: Rodriguez Celis, Esperanza
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: Chemistry -- Dissertations, Academic -- UF
Genre: Chemistry thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Material analysis and characterization can provide important information as evidence in legal proceedings. Taking advantage of the multi-element detection capability and minimal to no sample preparation of LIBS, this technique is proposed as a viable alternative for glass analysis. Discrimination studies by linear and rank correlation methods were performed to glasses from automobile windows. Linear correlation combined with the use of a spectral mask, which eliminates some high-intensity emission lines from the major elements present in glass, provides effective identification and discrimination at a 95% confidence level. In the course of this study, several key instrumental parameters were identified and investigated, and a comparative study regarding the performance of up to four different commercial instruments for LIBS analysis of solids has been made. Results indicate that the spectral resolution and sensitivity of the systems were linked to the performance and suitability of the technique for material identification by correlation methods, especially when samples were of very similar composition. While the qualitative characterization of materials without sample preparation is certainly one of the main advantages of LIBS, the possibility of performing accurate quantitative analysis relies on the use of calibration curves made with matrix-matched standards. Such quantitative analysis would also improve the discrimination capability of the technique in the case of the glass samples analyzed. Quantitation is one of the analytical aspects of LIBS where improvements are needed. In addition to the conventional analytical procedures followed in LIBS analysis, a normalization procedure based upon the possible correlation existing between the background fluctuation and the analytical signal. This method has been applied to several types of solid samples. The results of all these investigations have allowed a fair assessment of the applicability of LIBS for discrimination and quantitative characterization of solid samples in general, and glasses in particular.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Esperanza Rodriguez Celis.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Omenetto, Nicolo.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2009-11-30

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024353:00001


This item has the following downloads:


Full Text

PAGE 1

LASER INDUCE D BREAKDOWN SPECTROSCOPY AS A TOOL FO R DISCRIMINATION AND QUANTITATIVE CHARACTERIZATION OF GLASS FOR FORENSIC APPLICATIONS By ESPERANZA MARIELA RODRIGUEZ CELIS A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009 1

PAGE 2

2009 Esperanza Mariela Rodriguez Celis 2

PAGE 3

To m y family, for their love, support, and encourag ement; to my best friend and love of my life, Oscar, for everything that you are and ev erything you have brought to my life 3

PAGE 4

ACKNOWL EDGMENTS I would like to express my sincere gratitude to my research advisor Prof. Nicolo Omenetto for his continued guidance and encouragement. His knowledge, patience and constant support have positively contributed to my formation as a scientist. I will not forget the many hours of discussion and how he enthusiastically explaine d to me hard to understand concepts. I would never be able to show my full gratitude to him for his friendship and advice. I would like to acknowledge Dr. James Wineford ner for his valuable research guidance and encouragement during my studies. His always pos itive attitude, during my presentations in group meetings and seminars, strongly contributed to shape my speaker skills and conquer my fears of presenting scientific data. I am also grateful to Dr. Benjamin Smith for all the invaluable help he has provided. Starting in late 2003, when I applied to graduate school, Ben has always been there offering continuous support and guidance. I appreciate all the insightful comments during group meetings and in the lab. I have much appreciati on for Dr. Igor Gornushkin for teaching me to perform correlation analysis and LIBS in general. He always welcomed my ideas and patiently explained his. His support has been continuous, no t only while in Gainesville but also from his new job in Germany. I would also like to thank Dr. Uwe Heitmann and Dr. Patricia Lucena, visitor scientists to the Wine fordner-Omenetto-Smith group, for sharing their knowledge and for their friendship. I thank all the previous and current members of the Winefordner-Omenetto-Smith research group, who have become like a second family to me over these years. Everyone brings some unique quality that makes the laboratory a great place to work. Some former group members that I cannot leave out my friends L ydia Breckenridge, Pamela Monter ola, and Akua Oppong-Anane. I miss our time together and all the fun moments we shared in and out the lab. I also want to directly thank Galan Moore for hi s support in this research by providing the glass standards and 4

PAGE 5

for patiently correspondi ng with m e through e-mail to answer my questions about glass. My most sincere gratitude to th e present group members, Jonathan Merten, Dan Shelby, Farzad Pakdel, Heh-Young Moon, and Andy Warren for unders tanding the frustrations involved in LIBS analysis and for sharing moments of joy ever y time I got a linear ca libration plot. No list of acknowledgments would be complete without thanking Jeanne Kara bly, Lori Clark and Antoinette Knight for keeping the paperw ork and degree requirements in line. I made many sweet friends in Gainesville. I regret that space and time will not permit me to acknowledge them all by name. Nevertheless, I am thankful for the role every one of them played in my life and studies. I especially tha nk Elaine Muther and Cathy Silas who have been like mothers to me in the absence of my fam ily. They have been great companions when I needed them, listening and advising whenever they could. I would never forget our Sunday nights dinner and domino tournaments. I also want to thank my Spanish-speaking friends: Frank, Aybi, Giovanni, Yeneire, Henry, Pa ula, Elizabeth, Oscar, Soleda d, Karen, Jorge, Sarah, Fedra and Daniel. The time spent with you was remarkably fruitful and memorable. Needless to say, I am grateful, as always, to my family for their patience, prayers and encouragement. To my parents, Eduardo and Lidia, your support and ex ample have made me who I am. I hope you will always be proud of me. To my sister, Betsabe, brother, Wilson and, brother-in-law Abraham, who have also shown me great support, I want to say I am proud of you and of being your sister. I love you all so muc h. To my husband, Oscar, for all the support and encouragement and most importantly for his ge nuine friendship and love. I hope and pray we live a long and happy life together. Finally, I would like to acknowle dge my God who has always been the rock that makes up my foundation, and the calm in the storm. 5

PAGE 6

TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........9 LIST OF FIGURES.......................................................................................................................10 ABSTRACT...................................................................................................................................13 CHAPTER 1 INTRODUCTION TO LASE R-INDUCED BREAKDOWN SPECTROSCOPY................15 Introduction................................................................................................................... ..........15 History of LIBS......................................................................................................................15 Characteristics of the LIBS Process.......................................................................................18 Instrumentation................................................................................................................ .......20 Laser Systems..................................................................................................................20 Spectral Resolution Devices............................................................................................21 Detectors..........................................................................................................................23 Advantages of LIBS............................................................................................................. ..23 Considerations in the Use of LIBS.........................................................................................24 Sample Homogeneity......................................................................................................24 Matrix Effects..................................................................................................................25 Sampling Geometry.........................................................................................................25 Safety...............................................................................................................................25 Conclusions.............................................................................................................................26 2 BRIEF OVERVIEW OF GL ASS AS FORENSIC EVIDENCE............................................29 Glass as Forensic Evidence....................................................................................................29 Glass as a Chem ical Matrix....................................................................................................3 0 Microscopic Techniques for Glass Examination....................................................................31 Elemental Analysis of Glass Fragments.................................................................................32 Atomic Sp ectroscopy.......................................................................................................33 X-Ray Methods...............................................................................................................35 Inorganic Mass Spectrometry..........................................................................................36 LIBS and Glass Analysis........................................................................................................37 Conclusions.............................................................................................................................39 3 CORRELATION ANALYSIS AND THE DISCRIMINATION OF GLASS FOR FORENSIC APPLICATIONS................................................................................................42 Introduction................................................................................................................... ..........42 Experimental................................................................................................................... ........42 6

PAGE 7

Sa mples............................................................................................................................42 Instrumentation and Data Acquisition.............................................................................43 Software...........................................................................................................................44 Results and Discussion......................................................................................................... ..44 Sampling Considerations.................................................................................................44 Sample Identification by Linear and Rank Correlation...................................................47 Conclusion..............................................................................................................................52 4 COMPARATIVE STUDY OF FOUR LIBS SYSTEMS FOR THE ANALYSIS AND IDENTIFICATION OF SOLIDS...........................................................................................58 Introduction................................................................................................................... ..........58 Experimental................................................................................................................... ........59 LIBS Setup......................................................................................................................59 Samples............................................................................................................................60 Methods...........................................................................................................................60 Results and Discussion......................................................................................................... ..60 Spectral Resolution..........................................................................................................60 Limits of Detection and Precision...................................................................................61 Material Identification by Linear Correlation.................................................................62 Conclusions.............................................................................................................................63 5 CALIBRATION CURVES FOR TH E QUANTITATIVE ANALYSIS OF GLASS............70 Introduction................................................................................................................... ..........70 Experimental................................................................................................................... ........75 Glass Standards...............................................................................................................75 Instrumentation................................................................................................................ 76 Results and Discussion......................................................................................................... ..77 Optimization of Experimental Conditions.......................................................................77 Construction of Calibration Plots....................................................................................81 Determination of strontium......................................................................................81 Determination of magnesium...................................................................................85 Determination of titanium........................................................................................85 Conclusions.............................................................................................................................86 6 NORMALIZATION OF THE SIGNAL FOR QUANTITATIVE LIBS ANALYSIS........111 Introduction................................................................................................................... ........111 Experimental................................................................................................................... ......114 Results and Discussion......................................................................................................... 115 Calibration plot for silicon in steel.........................................................................116 Calibration plots in aluminum alloys.....................................................................117 Correlation coefficient as a functi on of the detector delay time............................118 Conclusions...........................................................................................................................118 7 CONCLUSIONS AND FUT URE WORK...........................................................................140 7

PAGE 8

Summ ary and Concluding Remarks.....................................................................................140 Future Research Directions...................................................................................................142 LIST OF REFERENCES.............................................................................................................144 BIOGRAPHICAL SKETCH.......................................................................................................162 8

PAGE 9

LIST OF TABLES Table page 2-1 Composition of soda-l ima container glass (SRM 621)......................................................41 2-2 Characteristics of the instrumental methods for the elemental analysis of glass. .............41 3-1 Information of the glass samples....................................................................................... 56 3-2 Identification using linear and rank correlation.................................................................56 3-3 Detected similarities between samples using p-values obtained by Students T-test .......57 4-1 Technical characteristics of the LIBS spectrometers used in this study............................67 4-2 Elemental composition in percentages (%) for Cast Iron Standards ................................68 4-3 Limits of detection and preci sionobtained with cast iron standards..................................68 4-4 Correct Material Identifi cation (%) using linear correlation.............................................68 4-5 Correct Material Identification (%) using linear correlation fo r converted spectra...........69 5-1 Chemical composition of the stud ied glass standards from NIST (SRM).......................106 5-2 Trace elements present in the glass standards NIST (SRM) used in this study...............107 5-3 Chemical composition in glass standards designed to be representative of ancient glass..................................................................................................................................108 5-4 Percentage relative standard deviatio n variability with the nu mber of laser pulses .......109 5-5 List of spectral lines presen t in the vicinity of Sr I 460.73 nm........................................110 6-1 Chemical composition in aluminum alloys NRC-IMI.....................................................137 6-2 Chemical composition in aluminum alloys APEX Smelter Co.......................................137 6-3 Chemical composition in aluminum alloys BAM...........................................................138 6-4 Chemical composition in steel standards BAM...............................................................138 6-5 Spectral emission lines for the analysis of aluminum alloys...........................................138 6-6 Results for the calibration curves and the analysis of Si, Cr, Mg, Fe, Sn and Mn..........139 9

PAGE 10

LIST OF FI GURES Figure page 1-1 Diagram showing main events in the LIBS process..........................................................27 1-2 A schematic overview of the temporal history of a LIBS plasma.....................................28 1-3 Diagram of a typical laboratory LIBS apparatus...............................................................28 2-1 The glass-manufacturing process.......................................................................................4 0 3-1 Intensity dependence with the number of laser pulses.......................................................54 3-2 Percentage relative standard de viation (%RSD) vs. number of pulses on the same spot .......................................................................................................................... ..........54 3-3 Percentage relative standard devi ation (%RSD) vs. number of pulses on separate spots ......................................................................................................................... .........55 3-4 Spectral lines that were masked for the analysis...............................................................55 4-1 Experimental setup used to compare the four systems......................................................64 4-2 Comparison of spectral reso lution for the four spectrometers...........................................65 4-3 Calibration plots for th e Mn II ion line at 293.9 nm..........................................................66 5-1 SRM-NIST Series 112.3 (glasses in powder and solid forms)..........................................88 5-2 SRM-NIST Series 112.4 (trace elements, wafer form)......................................................88 5-3 Corning glass standards................................................................................................. ....89 5-4 Diagram of the experimental LI BS system used in the experiments.................................89 5-5 Zero-order plasma images obtained at different positions in standard cast iron 236.....90 5-6 Zero-order plasma images obtained at different positions in SRM 1831..........................90 5-7 Zero-order plasma images and thei r intensities (a.u) for the glass standards ...................91 5-8 Zero-order plasma image for glass C showing five regions fo r detector binning.............92 5-9 Binning effects on the intensity, background, %RSD, noise and signal-to-noise ratio ....93 5-10 Strontium peak intensity depende nce with the number of laser pulses ............................94 5-11 Strontium peak at 460.73 nm............................................................................................. 95 10

PAGE 11

5-12 Strontium peak at 407.77 nm.............................................................................................9 5 5-13 Optimization of the gate de lay for Sr atomic line at 460.73 nm........................................96 5-14 Optimization of the gate de lay for Sr ionic line at 407.77 nm...........................................96 5-15 Calibration plot for Sr atomic line at 460.73 nm...............................................................97 5-16 Calibration plot for Sr ionic line at 407.77 nm..................................................................98 5-17 Strontium (460.73 nm) peak intensity variation with the number of laser pulses ............99 5-18 Strontium (460.73 nm) peak intensity dependence with the number of laser pulses .....100 5-19 Standard deviation vs. wavelength around Sr I 460.73 nm for SRM 616.......................100 5-20 Standard deviation vs. wavelength around Sr II 407.77 nm for SRM 616......................101 5-21 Magnesium peak at 517.27 nm........................................................................................101 5-22 Calibration plot for Mg atomic line at 517.27 nm...........................................................102 5-23 Magnesium peak intensity dependence with the number of laser pulses for standard....103 5-24 Standard deviation vs. wa velength for Mg 517.27 nm in SRM 1411..............................103 5-25 Titanium peak at 336.12 nm............................................................................................104 5-26 Calibration plot for Ti ionic line at 336.12 nm................................................................104 5-27 Titanium peak intensity dependen ce with the number of laser pulses.............................105 5-28 Standard deviation vs. wa velength for Ti 336.12 nm in SRM 612..................................105 6-1 Spectrum of steel in th e wavelength range 246-256 nm..................................................120 6-2 Background(s) for Si I 251.61 nm in steel.......................................................................121 6-3 Linear correlation between Pi for Si at 251.61 nm in steel C1 and the background........122 6-4 Linear correlation between Bi and Bi,average......................................................................123 6-5 Fluctuations of intensity of Si at 251.61 nm in steel for 50 individual spectra...............124 6-6 Fluctuations of background intensity of Si at 251.61 nm in steel for 50 spectra.............124 6-7 Fluctuations of intensity of background minus an arbitrary constant value ...................124 6-8 Peak intensity as a function of the background in steel C2.............................................125 11

PAGE 12

6-9 Peak intensity as a function of the background in steel C6.............................................125 6-10 Calibration plot for Si at omic line at 251.61 nm in steel.................................................126 6-11 Plot of -1 against Si concentration in steel....................................................................126 6-12 Comparison of conventional and normali zed calibration plots for Cr at 301.52 nm ......127 6-13 Comparison of conventional and normalized calibration plots for Cr at 283.56 nm ......128 6-14 Comparison of conventional and normali zed calibration plots for Mg at 285.21 nm ....129 6-15 Comparison of conventional and normali zed calibration plots for Fe at 302.11 nm ......130 6-16 Comparison of conventional and normali zed calibration plots for Sn at 283.99 nm .....131 6-17 Comparison of conventional and normali zed calibration plots for Mn at 288.96 nm ....132 6-18 Comparison of conventional and normali zed calibration plots for Si at 288.16 nm ......133 6-19 Spectrum of Si at 298.76 nm in silicon wafer..................................................................134 6-20 Peak intensity as a function of the background for Si at different delay times...............135 6-21 Correlation coefficient as a function of the delay time....................................................136 12

PAGE 13

Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy LASER INDUCED BREAKDOWN SPECTROSCOPY AS A TOOL FO R DISCRIMINATION AND QUANTITATIVE CHARACTERIZATION OF GLASS FOR FORENSIC APPLICATIONS By Esperanza Mariela Rodriguez Celis May 2009 Chair: Nicolo Omenetto Major: Chemistry Material analysis and characterization can provide important information as evidence in legal proceedings. Taking advantage of the multi-element detection capability and minimal to no sample preparation of LIBS, this technique is proposed as a viable alternative for glass analysis. Discrimination studies by linear and rank correlat ion methods were performed to glasses from automobile windows. Linear correlation combin ed with the use of a spectral mask, which eliminates some high-intensity emission lines from the major elements present in glass, provides effective identification and discrimination at a 95% confidence level. In the course of this study, several key instrumental parameters were iden tified and investigated, and a comparative study regarding the performance of up to four different commercial inst ruments for LIBS analysis of solids has been made. Results indicate that the spectral resolution and sensitivity of the systems were linked to the performance and suitability of the technique for material identification by correlation methods, especially when samp les were of very similar composition. While the qualitative characterization of materials w ithout sample preparation is certainly one of the main advantages of LIBS, the possibi lity of performing accurate quantitative analysis relies on the use of calibration curves made with matrix-matched standards. Such quantitative 13

PAGE 14

analysis would also im prove the discrimination ca pability of the technique in the case of the glass samples analyzed. Quantitation is one of the analytical aspects of LIBS where improvements are needed. In addi tion to the conventional analyt ical procedures followed in LIBS analysis, a normalization procedure base d upon the possible corre lation existing between the background fluctuation and the analytical si gnal. This method has been applied to several types of solid samples. The result s of all these investigations have allowed a fair assessment of the applicability of LIBS for discrimination and qu antitative characterization of solid samples in general, and glasses in particular. 14

PAGE 15

CHAP TER 1 INTRODUCTION TO LASER-INDUCED BREAKDOWN SPECTROSCOPY Introduction Laser-induced breakdown spectroscopy (LIB S), also called laser-induced plasma spectroscopy (LIPS), laser spark spectroscopy (LSS), and laser optical emission spectroscopy (LOES) is a fast-growing atomic emission anal ytical technique that involves focusing a high power, short pulse laser (usually in the nanosecond range) on a samp le surface. A very energetic plasma that is rich in electrons, atoms and ions is formed. The plasma radi ation, characteristic of the elements present in the sample, is collected and analyzed. LIBS provides a simple, fast and direct method of elemental analysis. Indeed, soli d, liquid or gaseous materials can be analyzed with little-to-no sample preparation, and detec tion limits for solid samples in the part-permillion (ppm) range. In this chapter, a brief summary of the hi storical development of LIBS is presented together with an overview of the characteristics of the LIBS process. Each important component of a LIBS apparatus and the many advantages of the method along with its limitations were also discussed. History of LIBS The evolution of LIBS has been summarized and discussed in great detail in books[1-4] and several review articles[5-22], therefore, only the most significant events in the history of LIBS are included here. The laser-induced pl asma was observed in 1960 shortly after the discovery of the pulsed ruby laser[23-25]. Brench and Cross[26] first reported a laser plasma at the Xth Colloquium Spectroscopicum Internatio nale in 1962. They used the ruby laser to produce vapors from metallic and non-metallic materials. 15

PAGE 16

Early on, the laser was used prim arily, but not exclusively, as an ablation source with cross-excitation to provide the spectrum[10]. In 1963, the first an alytical application of laserinduced plasma for spectrochemi cal analysis of surfaces was presented by Debras-Guedon and Liodec [27]. A year later, Maker, Terhune and Savage [28] reported the first observation of optically-induced breakdown in a gas. In the same year, Runge, Minck and Bryan described the use of a pulsed ruby laser for direct spark exc itation on metals.[29] Late r, collection of the plasma emission at time intervals was achieved us ing a streak camera and rotating mirrors in the detection system [1, 12]. In the period from 1964 to 1967 the first in struments based on this technique were developed by VEB Carl Zeiss (Germany) and Jarrell-Ash (USA) [11] These instruments employed cross excitation, where the light for sp ectral analysis was generated by an auxiliary spark discharge and the laser was only used for ablation. The instruments were later discontinued because they could not contend in accuracy a nd precision with competi ng technologies at the time such as conventional spark spectroscopy, electrothermal atomization atomic absorption spectrometry (ATA-AAS) and inductively coupled plasma-atomic emission spectrometry (ICPAES) [1]. Other research directions continued to develop. In 1966, Young, Hercher and Yu [30] described the optical properties of laser-induced air sparks. In the 1970s, spectrochemists were mostly interested in using LIBS for direct ab lation, excitation, and observation of the spark, while physicists studied the breakdown in gase s[10]. Scott and Strash iem [12] studied timeresolved spectra obtained using Q-switche d and non-Q-switched lasers. In 1972, Felske, Hagenah and Laqua [31] described the analysis of steel by means of a Q-switched laser. Through this years, much of the LIBS research app eared in the Russian literature [32-36]. Other 16

PAGE 17

signif icant milestone in the development of LIBS constitutes the generation of plasma in water [37-39] and the spectrochemical analysis of aerosols [40-43]. During this period, the sample matrix was recognized as an affecting parameter of the LIBS signal [44-46]. The initial findings showed that physical and chemical characteristics of the samples play a si gnificant role in signal strength, and repeatability. Since the 1980s, there has been a renewe d interest in spectrochemical application of LIBS driven by its unique advantages, significant technological developments in the components (lasers, spectrographs, detector s) and emerging needs to perform measurements under conditions not feasible with conve ntional analytical techniques [2]. As LIBS advanced further into the last decade of the 20th century, both analytical and industrial research employing LIBS developed rapidly. The technique began to move out from the basic science laboratory into the real world of applications. LIBS configurations concentrated on developing rugged, robust and field-portable syst ems. Optical fibers we re built into LIBS systems with the purpose of carrying the spark light to the spectrometer [10]. Today, LIBS remains a very active field. The world-wide LIBS comm unity has established international meetings that include LIBS 2000 (Tirrenia, Italy), Euro-Mediterranean LIBS Symposium (EMSLIBS) 2001 (Cairo, Egypt), LIBS 2002 (Orlando, USA), EMSLIBS 2003 (Crete, Greece), LIBS 2004 (Malaga, Spain) EMSLIBS 2005 (Aachen, Germany), LIBS 2006 (Montreal, Canada), EMSLIBS 2007 (Paris, France), North American Symposium on Laser Induced Breakdown Spectroscopy (NASLIBS) 2007 (New Orleans, USA), and LIBS 2008 (Berlin, Germany). In addition there have been a multitude of LIBS sessions at Pittcon and FACSS meetings. The interest in LIBS is also evident in the exponentia l increase in the number of publications and patents rela ted to fundamentals and applications of this technique. 17

PAGE 18

Significant progress has been m ade on the dive rse and versatile applications of LIBS including process monitoring (for example, in the metallurgical and mining industries), environmental monitoring, biomedical and pha rmaceutical applications, forensic analysis, military and homeland security, space exploration, and diagnostics of archaeological objects. As continuous improvements in instrumentation and data analysis are developed, the future for LIBS now appears brighter than at any time during its recent history. Characteristics of the LIBS Process The principle of LIBS is similar to th at of conventional plasma-atomic emission spectroscopy (AES), such as inductively coupled plasma ICP-AE S, microwave-induced plasma (MIP-AES), direct-current plasma (DCP-AES), arcand spark-AES. What distinguishes LIBS from these other techniques is that the sample does not need to be transported into the plasma source; indeed the plasma is generated at the sample surface making it a simpler method because the ablation and excitation proce sses are carried out by the laser pulse in a single step [14]. In LIBS, the vaporizing and exciting plasma is produced by a high-power, short pulse laser (usually in the ns range) on a sample (solid, liquid or gas). Each firing of the laser, or single-shot, produces a single LIBS measurem ent. In practice, however, the signals from various laser plasmas are added or averaged to in crease accuracy and precision [2]. LIBS of solids can be considered as a set of complex processes: laser-interaction with the solid, ablation or removal of samp le particles, and breakdown or plasma formation. Briefly, there are two main steps leading to breakdown. First, there is a generation of a few free electrons which play a role of initial receptors of en ergy through three body collis ions with photons and neutrals. Second, there is aval anche ionization in the focal region caused by collisions, ionization, more electrons, and energy absorption [2]. A schematic representation of the main processes for a laser-induced plasma on a solid surface is shown in Figure 1-1 [21]. 18

PAGE 19

The process is initiated by reflection or ab sorption (Fig. 1-1a) of energy by the sam ple from the pulsed laser. At moderate irradiances (~106 W cm-2), the absorbed energy is rapidly converted into heat, resulting in melting and vaporization of small portio n of material into ionized gas, when the local temperature approaches the boiling point of the material. In Fig. 11b, the removal of particulate matte r from the surface leads to the formation of a vapor in front of the sample. When this vapor condenses as droplets of submicrometer size, it leads to scattering and absorption of the laser radiation inducing heating, ionization and plasma formation (Fig. 1-1c). Other mechanisms that follow these include fa st expansion of photoa blated-material (Fig. 1-1d), formation of polyatomic aggregates and cl usters and deposition of the ablated and molten material around the crater (Fi g. 1-1e-f). During the expansion phase, the plasma emits useful signals. As it cools and decays, the ions and el ectrons recombine to form neutrals, and even molecules. Energy is released th rough radiation and conduction [2]. Because the laser plasma is a pulsed source, th e resulting spectrum evolves rapidly in time. A schematic representation of th e temporal history of the lase r-induced plasma concerning the different predominant emitting species is illustrated in Fig. 1-2 [2]. At the beginning, a white light, or continuum, dominates the pl asma light. This light is caused by bremsstrahlung from German bremsen to brake and strahlung radiation, and the recomb ination radiation from the plasma as free electrons and ions recombine in the cooling plasma. If the light is integrated over the entire duration of the plasma, the continuum light could interfere with the detection of weaker emissions from minor and trace elements in the plasma. Therefore, another important parameter in LIBS is time resolution; time-reso lved spectroscopy is essential to improve the sensitivity and selectivity in LIBS experiments because it allows rejecti on of the strong spectral 19

PAGE 20

continuum emitted at the beginning of the ablation [47]. The parameters used for time-resolved detection are td also known as time delay, the time between plasma formation and the start of the observation of the plasma light, and tb, or gate width, the time period over which the light is recorded. These parameters are highly dependent on the element and the matrix, and must be optimized for each type of sample In the literature, delay times in the range of 1-3 s, and integration times of 1-10 s are mostly reported [21]. Instrumentation The instrumentation for LIBS generally consists of a pulsed laser beam for sample ablation or breakdown, the optics for focusing the laser b eam and plasma emission, a spectral resolution device for wavelength selection and a detector. A typical LIBS set-up is shown in Figure 1-3. The basic components of any LIBS system are similar but they are tailored to the particular application. Laser Systems Intensity, directionality, monoc hromaticity, and coherence ar e the main properties that distinguish laser light from conventional light sources. Laser radiation can be emitted continuously or in short pulses, and even be tunable over a wide rang e of wavelengths [4]. Pulsed lasers are mostly used in LIBS and must generate pulses of sufficient power to produce the plasma. Besides the laser power it is also of importance the la ser ability to deliver the energy to a specific location. Th e power per unit area that can be deli vered to the target is known as irradiance and is also calle d flux or flux density. Since the discovery of the ruby laser [26], LI BS developments have been marked by the progress in laser technology. Each laser has its ow n properties such as wavelength, mode quality, and characteristics of operati on. The most widely used lasers for LIBS include the Nd: YAG, ruby, gas and excimer lasers. 20

PAGE 21

The Nd:YAG laser, with output wavelengths ( ) of 1064 nm, 532 nm (frequency-doubled), 355 nm (frequency tripled) and 266 nm (freque ncy quadrupoled) pulse length of 3-10 ns.[14] Nd: YAG lasers (flashlamp pumped) are preferre d for most LIBS applications because they provide a reliable, compact, and easy to use source of laser pulses togeth er with high irradiances [1]. The ruby laser with a wavelength ( ) = 694 nm and a pulse length of 10 ps (ruby picosecond pulse) 10 ms (ruby normal pulse) [4]. Gas lasers, including CO2 ( = 10.6 m and pulse length = 0.2 100 s) and N2 ( = 337 nm and pulse length = 3 10 ns) [4, 48]. These lase rs require periodic ch ange of gases, and do not couple well into many metals. Excimer lasers with pulse lengt hs of 10-35 ns, including ArF ( = 193 nm), KrF ( = 248 nm) and XeCl ( = 308 nm). They also require periodic ch ange of gases or gas flow and provide UV wavelengths only [1]. These lasers produce tipically-t ens-to hundreds of mJ per pulse and peak power outputs in the mW order. Once the laser pulses are focused by an appropriate lens to submicron-sized spots, the resulting irradiance is on the order of 1010 1012 W cm-2 which is enough to produce a plasma on solid samples. Spectral Resolution Devices The groundwork of a LIBS measurement is th e collection and analysis of an emission spectrum. Resolution and width of the spectrum that can be observed are important properties of a spectrometer. These specifications depend on th e particular applicati on in mind; however a larger spectral window is needed for multi-elem ent analysis. The following are examples of the most common spectral components of a LIBS system: Acousto-optic tunable filter (AOTF) 21

PAGE 22

Grating m onochromator Grating spectrograph Echelle spectrograph An AOTF is a diffraction based optical-band-pass filter that can be rapidly tuned to pass various wavelengths of light by varying the ra dio-frequency of an ac oustic wave propagating through an anisotropic crystal medium (e.g. TeO2). The transmitted light is then detected using a photon detector device [49]. In addition to a fast wavelength shifting, AOTF provide high energy throughput, robustness (no-moving parts) and usefulness for imaging [50]. A grating monochromator is a spectrometer that is tuned to monitor a selected wavelength which is presented at the exit slit of the device for detection. The most po pular design of grating monochromator is the Czerny-Turner monochromator. It provides high resolution with an f # ~ 4. A spectrograph is similar in basic configuration to a monoch romator except it has an exit plane at which a continuous range of wavelengths is presented for detection using some type of array detector or a series of single-wavelength detectors positioned behind individual slits. The spectral range recorded is limited by the width of the focal plane and the size of the array detector. However, it provides a wide spectral co verage with high resolution and f # ~ 4 [1]. For LIBS, echelle spectrographs are typically use d. For analysis of a wide range of samples, a system based on an echelle spectr ograph offers a combination of high resolving power ( / = 2500 10 000) and wide wavelength coverage (190-800 nm). The strong emission lines of most elements lie in this region. An echelle gr ating with a prism order sorter disperses the spectrum into two dimensions (wavel ength vs. order). Therefore, multiple exit slits and photomultipliers can be placed in the two-dimensional focal plane with f # typically >9. 22

PAGE 23

Detectors The choice of detector f or LIBS experime nts is dependant on the method of spectral selection and the number of elements to be m onitored. The simplest detectors available are photodiodes (PD) and photomultiplier tubes (PMT). These are highly sensitive devices that measure instantaneous light intens ity and are used together with line filters or monochromators providing only single wavelength detection. By placing many small photosensitive elements (pixels) in an array at the exit slit of the spectrometer a wider spectr al range can be cover. Examples include photodiode arrays (PDA), char ge-injection devices (CID) and charge-coupled devices (CCD). These are light-integrating devices since they collect th e light for a period of time, and then the signals stored on each pixel are read out sequentiall y, one pixel at a time further increasing the response time. For time-reso lved LIBS studies with array detectors, a micro-channel plate (MCP) is coupled to the detection system. The MC P acts as a regulator when the incident radiation is allowed to reach the detection device, amplifying the light by converting it to electrons which ar e then reconverted back to light before detection by the array detector. The coupling of a MCP with array de vices are known as intensified detectors (e.g. ICCD) [2] Advantages of LIBS LIBS, like other methods of AES, is able to detect all elements and has the ability to provide simultaneous multi-element detection capability with low absolute detection limits. In addition, because the laser spark uses focused optical radiation to form the plasma, LIBS exhibits numerous appealing features th at distinguish it from more conve ntional AES-based analytical techniques [2]. These are: simple and rapid or real time analysis, th e ablation and excitation processes are carried out in a single step; little-to -no sample preparation, which results in increased throughput and reduction of tedious and time-consuming sample digestion and 23

PAGE 24

preparation procedures (this, however, can lead to a loss of accuracy thro ugh contam ination).[13] LIBS allows in situ analysis requiring only optical acce ss to the sample. It can also be performed over a great distance, a t echnique referred to as remote sensing. Unlike remote analysis, in which some part of the LIBS system is close to the sample, is the method of stand-off analysis. Here, the laser pulse is focused onto the sample at a distance using a long focal length optical system [2]. Virtually any kind of sample can be analyzed: solids, liquids, aerosols, or gases. LIBS has the ability to analyze extremely hard materials wh ich are difficult to digest such as ceramics, glasses and superconductors [14]. It is a non-de structive method, very small amount of sample (~0.1 g 0.1 mg) is vaporized. It provides good sensitivity for some elements (e.g. Cl, F) difficult to monitor with conventional AES met hods. In addition, LIBS has adaptability to a variety of different measurement scenarios, e.g. underwater analysis, direct and remote analysis, compact probe with the use of miniature solid state lasers, stand-off analysis. Considerations in the Use of LIBS Among a few disadvantages of LIBS are the poor precisi on (5-15%), poor relative detection limits (in the ppm range), and matrix and spectral interferen ces. Sample homogeneity, physical and chemical matrix effects, sampling ge ometry, and safety in the analysis are some important considerations in the use of LIBS. Sample Homogeneity This is a direct consequence of no sample pr eparation and mostly affects the analysis of solids. Preparation typically produces a homogeneous sample ; the lack of preparation complicates the analysis of nonhomogeneous samples with a point detection method such as LIBS. The area interrogated is small, typically 0.1 1 mm diameter [2], involving a very small mass of material. Non-uniformities may be averaged out using a number of laser plasmas to 24

PAGE 25

repetitively interrog ate different areas of a sample and the results are combined to produce an average measurement. Matrix Effects Physical properties and sample composition affect the element signal such that changes in concentration of one or more of the elements fo rming the matrix alter an element signal even though the element concentration remains constant These effects complicate the construction of calibration plots and hence the abil ity to obtain quantitative results Matrix effects can be divided into two kinds, physical and chemical. Differenc es between specific heat heat of vaporization, thermal conductivity, absorption at the laser wavelength, and part icle size contribute to the presence of physical matrix effects. Chemical matrix effects occur when the presence of one element have an effect on the emission of anothe r element. For example, easily ionizable species increase the electron density, ther eby reducing the intensity of le ss ionizable components [1, 2]. Sampling Geometry When the power density is high enough, a plasma is formed on the surface of a solid even though the distance between the sample and the lens may be different from the lens focal length. The lens-to-sample distance affects the mass abla ted that, in turn, affects the element emission intensity. Maintaining the sample geometry constant is critical to obtain the best analytical results. Safety There are certain operational parame ters that must be consider fo r the safe use of LIBS [2]. These are: Ocular hazard by the laser High voltage circuits used on the laser operation Explosive potential of the laser spark for certain materials Possibility of generating t oxic airborne particles. 25

PAGE 26

Industry safety standards and preparation of Standard Operating Procedures (SOP) or Hazard Control Plans (HCP) that id entifies the hazards associated with the experiments and indicates steps to be taken to mitigate the hazards are recommended for LIBS users. Conclusions The development of LIBS has progressed ra pidly during the last decade. Numerous research groups have worked on improving LIBS measurements of different samples by using advanced lasers, detection systems and data processing methods. These improvements are opening new application areas for LIBS. However, there are still major difficulties in this technique. To overcome the known problems and advance LIBS for future applications, the knowledge of this technique needs to be expand ed through basic research in the fundamentals, quantification, pulse to pulse fl uctuations, and study of the la ser-material interactions[4]. 26

PAGE 27

sample Laser pulseab cd efabsorption r e f l e c t i o n crater vaporf r a g m e n t a t i o nf u s i o n m e l t i n gs u b l i m a t i o n a t o mi z a t i o n Shock wave absorption b r e m s s t r a h l u n g e m i s s i o nr a d i a t i o ne x p a n s i o n Figure 1-1. Diagram showing main events in the LIBS process: (a ) laser-material interaction, (b) heating and breakdown, (c) expansion and shockwave formation, (d) emission, (e) cooling and (f) crater forma tion. Adapted from Ref [21] 27

PAGE 28

1 ns10 ns100 ns 1 s 10 s 100 s Strong continuum emission Ions Neutrals MoleculesEmission Intensity Laser pulse Plasma Continuum tdtbElapsed time after pulse incident on target Figure 1-2. A schematic overview of the tempor al history of a LIBS plasma during which emissions from different species predominate. The box represents the time which the plasma light is monitored using a gatable detector; td is the delay time and tb the gate pulse width. Adapted from Ref [1, 2] Laser Focusing Optics Focusing Optics Sample Pulsed laser Wavelength Selector Detector LIBS spectrum Figure 1-3. Diagram of a typical laboratory LIBS apparatus. 28

PAGE 29

CHAP TER 2 BRIEF OVERVIEW OF GLASS AS FORENSIC EVIDENCE At identifying glass as forensic evidence, th ere is a continued move away from dependence on physical properties measured such as index of refraction and densit y, towards methods of elemental analysis of its trace components. This chapter focuses on the potential of LIBS for the discrimination of glass fragments for forensic applications. Glass as Forensic Evidence Trace evidence is a generic term for small, often microscopic, materials transferred between people, places, and objec ts. There is an enormous range of materials cover by this terminology including fibers, paint, glass, hair, soil, feathers, metal, brick, dust, sand, pollen, sawdust, and vegetation [51]. Fragments of brok en glass collected at crime scenes such as burglaries, car crashes, hit and runs, and vandalism constitute forensic trace evidence in criminal investigations[52]. These glass fragments, known as control sample s, can be compared with those recovered from the victims body and/or the suspects belong ings. If they are found to come from the same source, they might associate a suspect with the pe rpetration of a particular crime. Hence, it is essential that the method chosen for the analysis is capable of handling small sample fragments to provide adequate confidence in the results. Th ere are two main goals fo r the analysis of glass for forensic purposes [53]. First, to classify the glass fragments in to one of a number of possible categories, e.g. sheet, containe r, vehicle window, vehicle headlamp or tableware. This classification could help crime i nvestigators focus their search fo r an appropriate type of glass sample. Second, to determine whether two groups of glass fragments, the control sample, and the recovered sample from the suspect, could have shared a common origin. 29

PAGE 30

Glass as a Chemical Matrix Glass is defined as an inorganic product of fusion which has been cooled to a rigid condition without crystallization. Transparency, d u rability, electrical and thermal resistance, and a range of thermal expansions are a unique combination of glass prope rties [52]. The main component of glass is sili ca sand (60-75%). Silica (SiO2) is a glass former with a melting point higher than 1700C, and while it can be made in to a specialized glass where high temperatures resistance is required, other components or additives are added to simplify its processing. Soda ash (sodium carbonate, Na2CO3, 12-16%) and potassium oxide (K2O) are included in the mixture to reduce the fusion temperature. Lime or limestone (calcium oxide, CaO, 7-14%) or dolomite (magnesium oxide, MgO, <1%) are added as st abilizers to provide fo r a better chemical durability. Depending on the end use of the produc t and on the manufacturing process, additional ingredients are intentionally used. The following examples serve to illustrate the range of commercial glass compositions: Fused silica glass Soda lime silicate glass Borate silicate glass Lead glass Aluminosilicate glass Alkali barium silicate glass Borate glass Phosphate glass Chalcogenide glass Most glass manufacturers rely on a steady s upply of recycled cr ushed glass, known as "cullet," to supplement raw materials and to pr olong furnace life since cullet melts at a lower 30

PAGE 31

tem perature. Cullet also adds some measure of he terogeneity as trace contamination which is an ideal effect for the forensic sc ientist because it introduces add itional contrast between batches originating from the same plant. Heterogeneity is also imparted by iron or chromium in the sand deposits, contaminants of concern for the manufact ure process since they can lead to undesirable coloring. For decolorization, manufacturers rely on CeO2, As2O3, SbO, NaNO3, BaNO3, K2SO4, or BaSO4.Other contaminants such as potassium oxi des in soda ash, magnesium and aluminum oxides in lime, or even strontium in dolomite are useful for the forensic discrimination of glass [54]. Glass manufacture in all sectors broadly follows the stages illustrate d in Figure 2-1. Most manufactured glass has a specific soda-lime co mposition, producing windows (flat glass) for buildings and automobiles and containers of all types. Table 2-1 show s the typical composition of a soda-lime glass. The properties of these glasses make them suitable for a wide range of applications in container, sheet (or float, th e name of the process fo r the manufacture of most sheet glass), vehicle window (also made by the fl oat process), vehicle headlamp (a borosilicate glass) and tableware glass (including leaded glass). Microscopic Techniques for Glass Examination A range of techniques is available for the forens ic examination of glass. The possibility of a physical match between the fragments is explore d. This requires the two broken edges to match perfectly, an outcome that is hard to find in re al cases [55]. However, if the surface shape of a fragment is different fr om that of the control item, the frag ments could not have come from the same source. If the fragment shows fluorescent properties typical of a flat float glass and the control glass has no fluorescent properties, the gla sses are not from the same source either. Other physical properties such as colo r, thickness, refractive index (RI), and density are also examined. 31

PAGE 32

The m ethod of RI relies on the temperature vari ation effect first described by Winchell and Emmonds back in 1926 [56]. The RI of a liquid ch anges when heated and cooled while there is less RI variation for a solid. Ther efore, if a transparent glass fr agment is immersed in proper oil and then heated, the RI of the oil and the glass will be identical and the glass will not be seen. The temperature at which the glass disappears is recorded and by using standards a calibration plot can be drawn. The RI of an unknown gl ass fragment is determined by knowing the disappearance temperature [52]. The determination of RI is not a straightforward procedure; fragments from the same object might give a rang e of RI readings, making necessary the use of statistical tests such as t-test, cluster analysis to provide more incisive analysis of the data. The determination of the RI has been the technique of choice for many years [57-68]. Nevertheless, technological improvements in th e glass manufacturing process have led to less variability in physical and optical properties be tween manufacturers and different plants of the same manufacturer [59]. The re duction of the spread among RI values reduces the informing power of this technique and highlights the need for additional techniques to facilitate reliable identification. Elemental Analysis of Glass Fragments Ever since there is a better quality control of the batch components, the analysis of trace elements impurities within the raw materials c onstitute a useful path for discrimination. The following is a list of the most important factors ( not listed in specific order) forensic scientists consider when choosing an analytic al technique for analysis [69]. Detection limits Accuracy Precision Sample destruction 32

PAGE 33

Sa mple size requirements Total analysis time and ease of sample preparation Feasibility to control matrix effects and interferences Sources of error understood and controlled Linear concentration range Cost of equipment, operator expertise, and maintenance Level of reliable identification Several elemental analysis techniques are available for the characterization and discrimination of glass fragments. These available methods include atomic absorption and atomic emission spectroscopy, X-ray met hods, and inorganic mass spectrometry. Atomic Spectroscopy Atomic absorption [70] and atomic emission (AE) are two types of atomic spectroscopy which have been applied to the analysis of fo rensic glass samples. Flame Atomic Absorption Spectroscopy (FAAS), is simple to use, relative ly inexpensive, and provides outstanding sample throughput for the analysis of a small number of elements. Sensitivity in the part-per-billion (ppb) results in a minimal sample size required for analysis, generally 100 g or less. FAAS equipment is available in many forensic scie nce laboratories which perform gunshot residue analysis. Despite these features, the need for a different lamp per element, limited number of elements that can be analyzed, sample destruct ion and tiresome preparat ion makes the technique inflexible for forensic work. Analytical proce dures for the measurement of Mg, Mn, Fe, Cr, Na, and As in glass have been reported using AA [52, 69, 71]. In forensic glass analysis, three types of sources for AE have been reported: spark, inductively coupled plasma (ICP) and lasers. Using the spark s ource and AE, levels of Al, Ba, Ca, Fe, Mg, and Mn have been determined in glass samples weighing approximately 1 mg [72]. 33

PAGE 34

The developm ent of ICP-AE represented a significant advance in the use of emission spectroscopy and in the an alysis of glass for forensic purpos es [61, 73-80]. Glass samples can be introduced in the plasma either by dissolving them and nebulizing the resulting solution or by direct solid sampling. The initia l ICP-AES methods for glass analysis were primarily designed for purposes of classification. In 1981, an analyt ical method using this technique was developed to determine the concentration of Mn, Fe, Mg, Al, and Ba, and together with refractive index measurements, 91% of correct iden tification was achieved [75]. Over the next several years, the concentrations of additional elements were determined. The most widely used protocol for casework was developed for determining the concen trations of 10 elements (Al, Ba, Ca, Fe, Mg, Mn, Na, Ti, Sr, and Zr) with good precision in m illigram-sized glass fragments [61, 76]. ICPAES has also been used by the Food and Drug Ad ministration laboratories (FDA) to associate baby food containers to the manufacturing plants in which they were made and to identify sources of contaminated glass in cas es involving product tampering [80]. In ICP-AES, sample dissolution is a limitation because it destroys the sample, and it is time consuming compared to other methods. In a ddition, ICP-AES instrument ation is costly and requires more extensive operator training, r eason why only a limited number of forensic applications for this technique have been found. An alternative to sample dissolution is direct solid sampling. Laser ablation (LA) ICP-AES ha s the clear advantage of providing localized analysis with little-to-no sample preparation. The effects of th e laser parameters on the amount of glass ablated, and the analytical figures of merit of LA -ICP-AES for the study of glass have been reported in the literature [ 73, 74]. Another AE technique with direct solid sampling is LIBS. More on LIBS and forensic glass analysis wi ll be discussed further in this chapter. 34

PAGE 35

X-Ray Methods Num erous studies of discrimination and categ orization of glass fragments based on X-ray methods and direct comparisons with other elemen tal analysis techniques have been reported in the literature. X-ray analysis techniques su ch as Wavelength-Dispersive X-ray Fluorescence (WD-XRF)[81-83], Wavelength-Di spersive Electron Probe Micr oanalysis (WD-EPMA)[82, 84], Scanning Electron Microscopy w ith Energy-Dispersive X-ray microanalysis, (SEM-EDX)[82, 85, 86], Energy-Dispersive X-ray fluorescence spectrometry (ED-XRF)[61, 81, 87-89], synchrotron radiation X-ray fluor escence spectrometry [67, 9093] and Total reflection X-ray fluorescence (TXRF) [94] have been successfully applied for the analysis of forensic glass fragments. For example, SEM-EDX has been repo rted for discrimination of glass samples. In this study, 81 samples, indistinguishable by RI and density measurements, were efficiently discriminated (~97.5% correct identification) by SEM-EDX using calcium concentrations and the elemental ratios to calcium for Ti, Mn, Fe, Cu, Zn, As, Rb, Sr, and Zr [89]. Another SEMEDX method, with similar results, was later re ported using instead Na/Mg, Na/Al, Mg/Al, Ca/Na, and Ca/K concentration ratios [86]. The ability to standardize th e peak area ratios from different X-ray fluorescence instru ments for the collected glass data to be related and compared has also been studied [88]. Effective glass id entification of 23 glass fragments was achieved by comparison of five elemental rati os of Ti/Sr, Mn/Sr, Zn/Sr, Rb /Sr, and Pb/Sr calculated from TXRF spectra [94]. Synchrotr on radiation XRF has also been successfully applied to the discrimination of glass using quantitative data of six elements: Ca, Fe, Sr, Zr, Ba, and Ce [93]. The most significative advantage of X-ray methods is that they are non-destructive. In addition, spectra are relatively si mple, there is little spectral lin e interference, relatively small samples can be analyzed and the multi-elemen t analysis capability makes it a speedy and convenient technique for forensic samples. The main limitations are that matrix-matched multi35

PAGE 36

elem ent standards are needed to obtain accurate quantitative results. Actually, quantitative analysis of forensic glass has been best achieved by an evaluation of the ratios rather than by the absolute concentration of the elements. Very small and irregularly sh aped samples are not amenable to this type of analysis. Commonly, sample preparation requi res embedding the glass fragments in a plastic resin and then polishi ng the surface until flat by grinding methods. The surface is usually coated with a carbon layer and the fragment is sampled at different locations [52]. Inorganic Mass Spectrometry Inductively coupled plasma mass spectrome try (ICP-MS) combines the multi-element capability and the broad dynamic range of ICP emi ssion with the enhanced sensitivity and ability to perform quantitative analysis of the elemen tal isotopic concentrati on and ratios. Typically, samples are introduced into an ICP-MS by aspi rating a solution of the sample. Often, liquid samples require little preparation, but solid samp les need to be dissolv ed. This preparation process requires time and use of acid-dissolution reagents. ICP-MS has been successfully applied for the discrimination of glass [68, 70, 92, 95-103]. The first reported application of ICP-MS to fore nsic glass analysis was made in 1990 [70] Seven glasses, statistically indisti nguishable by RI, were 85-90% successfully discriminated. The samples were as small as 500 g with detection limits below 0.1 ng mL-1. Later, a more extensive investigation of ICP-MS analysis was pres ented [101]. Sample di gestion methods were compared and up to 62 elements were determined in a range of glass samples. Successful differentiation of glasses of similar RI was acco mplished by comparing element concentrations and ratios (e.g., Sr/Ba). The precision of ICP-MS percentage relative standard deviation (RSD) < 3.9%, in trace element concentration determin ation was sufficient to provide adequate discrimination. 36

PAGE 37

The incorporation of laser ablation (LA) in IC P-MS has greatly sim plified the analysis of glass samples [100, 103-120]. There is already a va lidated method for glass discrimination using LA-ICP-MS [107]. The elemental menu comprises 10 elements: K, Ti, Mn, Rb, Sr, Zr, Ba, La, Ce, and Pb. It was shown than the method could be used for glass fragments sizes down to 1mm2 with limits of detection (LOD) in the order of g g-1, and precision and accuracy <10% for most of the measured elements. Current work suggest s LA-ICP-MS offers great promise for the fast and accurate multi-element comparison of sma ll samples in a non-destructive manner. Advantages of this technique include mi nimal sample preparation, multi-elemental capability, greater sensitivity and better det ection limits than conve ntional absorption and emission techniques, speed of analysis, and mi nimal sample destruction and contamination. However, in spite of its relatively high sensit ivity, this technique is very expensive, which precludes its use in many forensic laboratories. Another limitation, which also affects LIBS, is its matrix dependence: laser parameters cha nge depending on the matrix. Moreover, the quantification is less straightforward than with solution analysis due to the lack of solid calibration standards, particularly matrix-matched standards. Laser ablation is also susceptible to elemental fractionation. Fractionation is a dynami c process that include s the effects of the ablation, sampling, transport, and ionization and is defined as pr oducing ablation products that are not stoichiometrically representa tive of the sample composition [52, 115]. LIBS and Glass Analysis In this study, LIBS is proposed as a viable alternative for glass an alysis. LIBS has the potential to become an attractive technique for fo rensic applications. A fe w forensic applications have been reported for the analysis of gun pulse residues[121], minerals in hair[122], Rb traces in blood [123], detection of latent fingerprints [124], wood in a murder case [125], analysis of human cremation remains and elemental compositi on analysis of prosth etic implants [126]. 37

PAGE 38

Our group has evaluated the potential of LIBS for discrim ination of glass samples of similar RI values by comparing the LIBS spectra over a short period of time (same day) [127]. Research by Bridge et al.[108] have recently fo cused on the characterization of automobile glass fragments by LIBS and LA-ICP-MS. For the LIBS analysis, 18 ionic and atomic emission lines, from the elements Al, Ba, Ca, Cr, Fe, Mg, Na, S n, Si, and Ti, were evalua ted and used to form ten different ratios of emissi on intensities. With use of th ese ratios, 93 and 82% correct discrimination of 23 glass samples was achie ved at confidence intervals of 90 and 99%, respectively. With the addition of RI data, the discrimina tion was improved to 100 and 99% for the confidence intervals of 90 and 99%. Later, this study was extende d to the examination of four sets of glasses (automobile windows, headlamp, and side mirror, and dr ink container glasses). The use of LIBS in combination with RI dete rmination provided 87% discrimination at a 95% confidence level[128]. Characterization of the influence of irradiat ion wavelength has been carried out on the analytical results form glass matrices with va rying optical and elemental composition by Barnett et al.[129] Two harmonics of the Nd: YAG laser (266 and 532 nm) were used to create the plasma of several glass standards and soda -lime glass samples. Good correlation for the quantitative analysis results for Sr, Ba, and Al were reported along with the calibration curves. The Nd: YAG 532 nm laser produced greater emissi on intensities with less mass removal than the 266 nm laser. Later, the authors present a comparison study of LA-ICP-MS, micro-XRF, and LIBS for the discrimination of 41 automotive gl ass fragments [130]. Excellent discrimination (>99%) results were obtained for each of the methods. Different combinations of 10 element ratios for the elements Al, Na, K, Ca, Fe and Sr were employed for the analysis and discrimination by LIBS. 38

PAGE 39

Consequently, good perform ance of LIBS is encouraging for its use in forensic laboratories. There have been pr evious studies of glass analys is by LIBS, although they did not focus on discrimination analysis for forensic purposes [131-154]. Conclusions Currently, there are a number of satisfact ory techniques available for the elemental analysis of glass fragments for forensic purposes Each of the various instrumental techniques has particular strengths and limitations. Table 2-2 summarizes the characteristics of some of the instrumental techniques described in this chapter. Every one ha s its advantages and disadvantages and all of them have found applic ations in forensic science laboratories. 39

PAGE 40

RAW MATERIALS Storage Weighing Mixing MELTING Refining Homogenization FORMING Shaping ANNEALING Controlled cooling SECONDARY PROCESSING Tempering Coating Coloring and discoloring Figure 2-1. The glass-manufacturing pr ocess. Adapted from Ref. [52] 40

PAGE 41

Table 2-1. Com position of soda-lima container glass. Standard Reference Material (SRM) 621 from the National Institute of Standards and Technology (NIST). Constituent Percent b y wei g ht SiO2 71.13 Na2O 12.74 CaO 10.71 Al2O3 2.76 K2O 2.01 MgO 0.27 SO3 0.13 BaO 0.12 Fe2O3 0.040 As2O3 0.030 TiO2 0.014 ZrO2 0.007 Table 2-2. Characteristics of the instrumental methods for the elementa l analysis of glass. Adapted from Refs.[127, 130] Characteristics AA XRF ICP-AES ICP-MS LA-ICP-MS LIBS Detection limit (ppm) 1 100 0.1 1 < 1 < 1 10-50 Sample penetration (microns) 100 80 50100 Multi-element No Yes Yes Yes Yes Yes Destructive Yes No Yes Yes No No Sample preparation Yes Yes (low) Yes Yes No No Cost Low Moderate Moderate High Very high Low Ease of use Easy IntermediateIntermediateDifficult Difficult Easy Glass discrimination Fair Good Good Very goodExcellent Good 41

PAGE 42

CHAP TER 3 CORRELATION ANALYSIS AND THE DISCRI MINATION OF GLASS FOR FORENSIC APPLICATIONS Introduction As discussed in chapter 2, material analysis and characterization can provide important information as evidence in lega l proceedings. The potential of LIBS for the discrimination of glass fragments based on correlation analysis is presented in this chapter. In this study, we examine the LIBS spectra of glass from a slightly different perspective th an the ones available in the literature [108, 128, 130]. That is, we do not seek a detaile d chemical composition or to calculate intensities or intensity ratios of some particular elements. Instead, we identify glass fragments from their unique LIBS spectral fingerp rints by using statistica l correlation methods. The procedure used in this research is the following. First, an unknown glass fragment is identified by correlating its spec tra against an availa ble spectral database. Second, the spectra of the fragments are compared against each other to statistically determine if they originated from the same glass source. Third, the long-term reproducibility of the analysis is presented. Optimal sampling conditions for acquisition of accurate LIBS spectra are also reported. A summary of the results and their statistical significance is presented. Experimental Samples A total of ten fragments from seven automobile glasses (side and rear windows) were used in this study. These ten fragments provide 45 possible pair comparisons. They were collected from automobiles at a local au to glass shop corresponding to 5 y ears of manufacturing and five vehicle manufacturers. Their height, length, and width were approximately 0.3, 1.0, and 0.9 cm, respectively. All of them were transparent and un coated. A more detailed sample description is shown in Table 3-1. 42

PAGE 43

All sam ples were mounted on a microscope glass slide using doublesided mounting tape and then placed on an XYZ translation stage that allows movement of the sample to a fresh spot. A laser positioning system consisting of a diode laser and a photodiode detector ensured a reproducible position of the sample. Instrumentation and Data Acquisition The LIBS instrumentation used in this st udy consisted of an O cean Optics (Dunedin, FL, USA) LIBS2000+spectrometer coupled to a Big Sky (Bozeman, MT, USA) Ultra Q -switched Nd:YAG laser operating at 1,064 nm. This laser delivered a maximum of 50 mJ in 10 ns, providing an irradiance of approximately 5.2 GW/cm2 on the sample surface. The laser beam was focused onto the glass surface using a 5-cm focal length lens. The radiation emitted by atoms ablated from the glass samples was collect ed by a quartz lens and guided into a sevenchannel spectrometer. It is worth pointing out that the optical collec tion system used in this work was not achromatic: this may affect the line inte nsities obtained in widely different spectral regions and as a consequence also the discrimi nation power. Chromatic effects should then be taken into consideration when transferring differe nt spectra from laboratory to laboratory. Each channel covered a spectral range of about 100 nm; the full range of the spectrometer was from 200 to 980 nm. The detector was a linear CCD with 2,048 pixels. The instrument spectral resolution (full width at ha lf maximum) was 0.1 nm. The instrumental parameters used were as follows: laser power 50 mJ per pulse, detector delay time and gate width 1 s and 2 ms, respectively. For data acquisition, each sample was ablated at 15 positions; each position consisted of 130 ablation pulses at 1 Hz and the data were obtained at atmospheric pressure. The first 30 sp ectra were discarded and the next 100 were averaged, providing an individual spectrum pe r position on the sample For the correlation analysis, 15 individual spectra per sample were averaged to create a sample library. 43

PAGE 44

Softw are A homemade program for correlation analysis written in Visual Basic 6.0 was used [155]. This program has already been successfully applie d to the identification of different classes of materials using their LIBS spectra [155-158]. The program allo ws creation of libraries by averaging individual spectra. Linear and rank co rrelation coefficients are calculated for each individual spectrum versus the library. Finally, a correlation plot is di splayed, corresponding to the maximum correlation coefficient and the name of the sample associated with the highest correlation coefficient. Results and Discussion Sampling Considerations Experimental conditions capable of providi ng high precision and repeatability between experiments are needed to obtain an accurate spectral fingerprint of each sample. First, the effect of the laser power was studied. The Nd:YAG laser used has a specified maximum pulse energy of 50 mJ which can be changed in increments of 5 mJ. In our experiments, it was found that the highest laser power (50 mJ) resulted in better signal-to-noise ratios. In LIBS, small unpredictable experimental fluctuations can cause a significant change in the appearance of the spectra. To check the st ability of our measurement setup the following procedure is envisaged: the intensity of a Si atom ic line at 288.16 nm is chosen and monitored at regular time intervals in a standard glass sample (NIST 612) used as a reference for the optimization of the experimental setup. The analysis is not continued and correlation is not performed if the difference in the intensities observed for this line exceeded the expected statistical variation of our method, = %. In this case, the en tire experimental setup is inspected and optimized again until the results are satisfactory. In particul ar, three experimental 44

PAGE 45

param eters are checked, namely, the alignment of the collection optics with respect to the plasma volume observed, the focusing distance of the laser beam on the target, which affects the ablation efficiency, and the short-term pulse-to-pul se energy fluctuatio n of the laser. A detector delay of 1 s and a fixed spectrometer integrat ion time of 2.1 ms were used. These values resulted from an optimization study carried out for the detection of carbon in soils[159]. Although the choice seems to be arbitrary, in view of the different matrices examined, the optimization study was repeated with different samples such as glass and cast iron standards, confirming the choice of the values obtained pr eviously. The repetition rate did not play a significant role during data acqui sition and was set at 1 Hz. No sample preparation is required for elementa l analysis by LIBS. However, relatively low spectral intensities were recorded after the first laser pulse wh ich slowly increased, reaching a constant value after approximately 25 pulses on th e same sample spot; this behavior is only observed when working with glass fragment sample s. Fig. 3-1 presents a typical plot of the variation of net intensities for the glass fragments (number of counts above the background level) for six emission lines versus the number of lase r pulses (up to 250) at one spot on the sample. This behavior is caused by the la serglass interaction [73, 106]. For the first few laser pulses, the samples were almost transparent and there was minimal ablation. However, as the glass interrogation progresses, defects in the glass we re formed, making ablati on stronger. Thus, about 25 preparation pulses were needed, in our case, to achieve reproducible ablation. The percentage relative standa rd deviation (%RSD) of the net intensity versus the number of laser pulses at one sample spot (in groups) is shown in Fig. 3-2. It is observed that the RSD values decrease from 100 to 300% RSD for first 30 pulses to 15% RSD for the next 200 pulses. Consequently, the first 30 ablation pulses were discarded and considered as preparation 45

PAGE 46

pulses. The influence of the roughness of the gl ass surface was not exam ined but it has been reported not to cause significant changes in the emission spectra of glass [120, 137]. To determine the number of individual spectr a to be averaged for the analysis, 180 pulse measurements were made at each position; ten positions per sample were examined. The first 30 spectra of each measurement were discarded and the remaining spectra were averaged in groups. The percentage RSD was calculated by using the averaged signals from each of ten sequential measurements. The plot of the percentage RSD against the number of averaged individual spectra is presented in Fig. 3-3 for some em ission lines. It was found the percentage RSD reached a constant level of 10% after approximately 30 pulses; therefore, 100 individual spectra were averaged per position in our measurements. Next, the effect of day-to-day changes in hum idity, which might affect the laser air spark [160], was studied. The intens ity of the hydrogen Balmer emission line at 656.3 nm was monitored for this purpose since its variation is an indication of the changes in humidity of the ambient air. No significant changes in humidity were observed in 1-week period, and to simplify the sampling our experiments were performed in ambient air. In general, the overall reproducibility of the spectra taken on di fferent days was within a variation of = %. The effect of the offset distance or focal de pth on pulse-to-pulse reproducibility was also considered. The offset distance is the difference between the lens-to-surface distance and the focal length of the lens[161]. It is an important parameter that mu st be held constant to obtain reproducible spectra[162]. No significant chan ge in precision was found when moving the sample up or down 5mm relative to the focal point of the lens and we simply focused the laser on the samples surface. 46

PAGE 47

The m inute amount of material removed per sample spot was approximately 3 g. The mass ablated per spot was determined after examining the crater made with a calibrated optical microscope and using the known value of th e density. The overall precision of these measurements was estimated to be approximately 10%. This mass of material has been proven to be representative of the whol e piece of glass in studies using synchrotron micro XRF spectrometry[163] a nd LA-ICP-MS[114]. Sample Identification by Linear and Rank Correlation Linear correlation measures the strength of the linear relation between two variables, in our case two glass LIBS spectra. The Pearson correlation coefficient, also known as linear correlation coefficient, is given by i i i i i i iyyxx yyxx r2 2)()( )()( (3-1) where x represents the mean of x s and yis the mean of the ys from two data sets; xi and yi are intensities of the two spectra measured at the same pixel i. We also used the rank (Spearman) correlation coefficient. The equation for rank correlation is similar to Eq. 3-1 but the values of the x and y distributions are replaced by their corresponding ranks R and S: i i i i i i iSSRR SSRR r2 2)()( )()( (3-2) where the ranks have numbers 1, 2, 3, N The highest rank, N is the total number of data points, or the highest pixel number. These ranks repl ace the true values of the x values and the y values in accordance with their magnitudes. 47

PAGE 48

The value of r lies b etween -1 and +1. Values close to zero indicated uncorrelated data sets. The absolute value of r can be used as an indicator of the association between the data sets since the strongest correlation is represented by the absolute value of 1. This study focused on the identification of ten glass fragments using linear and rank correlation coefficients. Fifteen individual spect ra were collected from each sample. Each spectrum was the average of 100 ablation pulses. We refer to a library as the collection of averaged spectra; a library was created containing ten averaged spectra (library 1). Each of the 150 individual spectra were co rrelated against the library spectra. The highest correlation coefficient indicated a similarity of a tested spectrum to one from the library. The difference between this and other correlati on coefficients indicates spectral and, hence, compositional differences. The results obtained on the yesno basis for the identification of the 150 individual spectra (set 1) using correlation coefficients are presented in th e top half of Table 3-2. These results include the name of the library the individual spectrum was identified as and the number of times (out of 15) corr ect identification was achieved. A ll the data points present in the spectra (13,701 pixels) were used for the correlation. From this ta ble, linear correlation suggests that there is similarity between samples A2 and A3 and between samples A5 and A6, which is in fact correct since those glass fragments came from the same window. However, linear correlation did not identify similarities between samples A8 and A9, which also came from the same source. The rank correlation also indicates si milarities for pairs A2 and A3 and A5 and A6, as does the linear correlation, but in addition suggests similarities for pair A1 and A9, which are not expected since these two samples came from different sources. Rank correlation also failed to identify similarities between samples A8 and A9. It is important to mention that all these ten 48

PAGE 49

sam ples have very similar LIBS spectra, and th e correlation coefficients between these samples are all in the range of 0.9 or higher. This is understandable since all the samples have very similar elemental composition. It is important that identifica tion is reproducible within an arbitrary long time period, e.g., spectra from unknown samples taken on different days could be compared with the existing spectral library. For this purpose, a new experime nt was performed in a period of 1week. This time, however, only five individu al spectra were collected from each sample, under the same experimental conditions reported earlier, and a new library was created containing ten averaged spectra (library 2). The identification of individua l spectra for this second set of data compared with library 2 showed similar re sults to the results obtained prev iously and presented in the top part of Table 3-2. In addition, this set of individual spectra was correlated versus the library created the previous week (library 1). The botto m half of Table 3-2 s hows that the correlation coefficients obtained by linear corr elation indicate similarities fo r pairs A2 and A3, A5 and A6, and A8 andA9, which came from the same window ; there were also no misidentifications. On the other hand, rank correlation also suggests si milarities for the three previous pairs but in addition shows similarities for pairs A1 andA9 a nd A4 and A5, which were not expected since the samples were from different sources. So far, it is felt that linear correlation provides good identification for samples that ar e the same and distinction for th e ones that are different even when the data are obtained on different days. Therefore, linear correlation is robust with respect to changes in spectral inte nsities on different days. Besides the apparent differences and similarities between samples determined by the use of correlation coefficients, strict statistical criteria must be applied in order to quantify the level of significance. To do so, a simple analysis by Students t test was applied. This hypothesis test 49

PAGE 50

determ ines whether two normally distributed populations are significantly different[164]. The normality of the distribution of the correlation coefficients was confirmed using normality plots ( Q Q plots). If the tests p value is less than the significance level chosen ( = 0.05 to give a 95% confidence level) the null hypothesis is rejected and it is concluded that the samples are different. Otherwise, the result s suggest there is no significant difference between the two populations; in other words, there is no signifi cant difference between the two samples. Based on these p values, the top half of Table 3-3 presents the discrimination between glass fragments when the second set of indivi dual spectra was correlated agains t the library created a week earlier (library 1). The p values for linear correlation coefficients indicate no significant difference for the means of pairs A2 and A3 and A5 and A6, which was expected. However, linear correlation also indicates that a significant difference exists between samples A8 and A9, which is not true. Therefore, linear corr elation provides 98% corr ect identification at a 95% confidence level. On the other hand, the p values for rank correlati on coefficients indicate no significant difference for the me ans of pairs A2 and A3, A5 and A6, and A8 and A9, which is correct and expected. However, this analysis also indicates similarities for pairs A1 and A8, A1 and A9, A1 and A10, and A4 and A5, which is incorrect, providing an overall 91% correct identification at a 95% confidence level. We conclude, therefore, that linear correlati on provides better results than rank correlation but does not find similarities for samples A8 a nd A9. Rank correlation finds these pair to be similar but suggests that four ot her pairs are also similar. To improve correlation analysis of spectra, m asking the spectra prior to evaluation was chosen. Masking is a simple multiplicative proces s that retains only the selected peaks of the components to be analyzed, thus eliminating analyt ically useless parts of the spectra (e.g., peaks 50

PAGE 51

or bands from the m atrix) [165, 166]. The criteria for choosing a valid mask is to strengthen spectral similarities between samples that are the same and improve differences for the ones that are different. One of the consequences of better qual ity control in the manuf acture of glass is less variability between concentrations of major el ements, e.g., Si, Mg, Na, Ca, Fe, and Al. As a result, analysis of trace element impurities with in major materials emerges as a useful path for discrimination [70]. Some high-intensity lines from major elements present in glass were blocked since the concentrations of those elements are expected to be very similar among the glass fragments. By doing so, we focused our anal ysis on elements of lower concentration. Fig. 4-4 shows the spectral lines that were masked for the analysis. It is advantageous to eliminate regions of th e spectrum where no lines are present since the noise affects negatively the correlation coeffici ents. Another possibility of masking is to correlate lines from the trace elements present in the samples instead of just blocking the highintensity lines; however, this might require a high er-resolution spectrometer capable of resolving and detecting K, Ti, Mn, Rb, Sr, Zr, Ba, La, Ce and Pb, which have been proven to provide effective discrimination by LA-ICP-MS [107]. The results obtained when using the spectral ma sk are presented in the bottom half of Table 3-3. Linear correlation re vealed that three pairs are indi stinguishable (A2 and A3, A5 and A6, and A8 and A9); therefore, linear correlation together wi th masking provides 100% correct identification at the 95% significance level. By ra nk correlation, these pairs are also similar, but three pairs (A1 and A9, A1 and A8, and A4 and A5) out of 45 possible pair combinations are also judged to be similar, yi elding 93% correct identification at a 95% significance level. It is clear that masking improves the result by linear correlation. Ho wever, there is no significant improvement with rank correlation, probably owing to the nature of the mask itself, 51

PAGE 52

which only blocks high-intens ity lines. Current research in our group suggests that rank corre lation is very sensitive to noise and less sensitive to small systematic changes in line intensities [167]. Therefor e, only linear corre lation is the method of choice. Conclusion In this work, linear and rank correlation tec hniques were applied for discrimination of LIBS spectra from glass samples with similar chemical composition, some of them from the same source. The robustness of this techni que was demonstrated by the 100% correct identification (95% confidence level for a type 1 error) obtained by linear correlation when used in combination with a spectral mask. The iden tification was reliable even when experiments were performed on different days when ambient c onditions might be different and affect the line intensities in the LIBS spectra. The rationale of using spectral masking is to eliminate regions of th e spectra containing several intense lines common to all samples and to take advantage of the trace element impurities present in these glasses. We are aware of the fact th at there are more sophisticated ways to generate a mask than the one used in this study. However, it was felt useful to focus at first on a simple masking procedure to see whether any fu rther elaboration of this concept was worth pursuing. More refined procedures are planned for the future. The main advantages of this LIBS method rela tive to other elementa l analysis studies of glass for forensic applications are: Lack of sample preparation Short analysis time, since sp ectral acquisition and correlation analysis could be done in minutes. No need for quantification or calculation of intensity ratios No need for supplemental measurements such as measuring RI values 52

PAGE 53

Future work for this study m ight include the use of masks for spectra l regions instead of lines, the use of background corre ction, and a study of different normalization procedures to assess the quantitative analytical performance of the technique. In conclusion, elemental analysis of glass by LIBS has the potential of becoming a useful technique for the discrimination of forensic glasses. Its usefulne ss as an analytical method for legal purposes will be determined by its general acceptance in the relevant scientific community. Evidence of general acceptance normally incl udes known error rates and publication of the methods in peer-reviewed journals[168]. These legal aspects and corresponding implications, which would require more in-depth statistical analysis, have not been considered in this work. 53

PAGE 54

11 01 0 0 0 200 400 600 800 1000 1200 1400 1600 Net intensity (counts)Number of laser pulses SiI288.2nm AlI309.3nm NaI330.2nm FeII259.9nm CaI527.0nm MgII279.6nm Figure 3-1. Intensity dependence with the number of laser pulses 1-1516-3031-5051-100101-150151-200201-250 0 50 100 150 200 250 300 % RSDNumber of laser pulses SiI288.2nm AlI309.3nm NaI330.2nm FeII259.9nm CaI527.0nm MgII279.6nm Figure 3-2. Percentage relative standard deviation (%RSD) variab ility with the number of laser pulses on the same spot 54

PAGE 55

11 01 0 0 0 5 10 15 20 25 30 35 40 % RSDNumber of laser pulses SiI288.2nm AIl309.3nm NaI330.2nm FeII259.9nm CaI527.0nm MgII279.6nm Figure 3-3. Percentage relative standard deviation (%RSD) variability with the number of laser pulses on separate spots 2002503003504004505005506006507007508008509000.0 0.2 0.4 0.6 0.8 1.0 Normalized Intensity (a.u)Wavelength (nm)A1 Mg Fe, Mg, Si, Ca Ca Fe, Mg Na Figure 3-4. Spectral lines that were masked for the analysis 55

PAGE 56

Table 3-1. Infor mation of the glass samples Sample name Vehicle information A1 2007 Buick Lucerne, side window A2 2002 Ford Focus, rear window A3 2002 Ford Focus, rear window A4 2003 Honda Accord, side window A5 2003 Toyota Tundra rear window A6 2003 Toyota Tundra rear window A7 2002 Toyota Camry side window A8 2004 Chevy Blazer side window A9 2004 Chevy Blazer side window A10 2003 Chevy Trailblazer side window Table 3-2. Identification using linear and rank correlation. First and second set of individual spectra are correlated against spectral library 1. A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 First set of individual spectra co rrelated against spectral library 1 Linear A1 (15/15) A2 (10/15) A3 (5/15) A3 (13/15) A2 (2/15) A4 (15/15) A5 (8/15) A6 (7/15) A6 (11/15) A5 (4/5) A7 (15/15) A8 (15/15) A9 (15/15) A10 (15/15) Rank A1 (12/15) A9 (3/15) A2 (12/15) A3 (3/15) A3 (14/15) A2 (1/15) A4 (15/15) A5 (11/15) A6 (4/15) A6 (8/15) A5 (7/15) A7 (15/15) A8 (15/15) A9 (13/15) A1 (2/15) A10 (15/15) Second set of individual spectra co rrelated against spectral library 1 Linear A1 (5/5) A2 (5/5) A3 (1/5) A2 (4/5) A4 (5/5) A5 (5/5) A6 (1/5) A5 (4/5) A7 (5/5) A9 (5/5) A9 (1/5) A8 (4/5) A10 (5/5) Rank A9 (5/5) A2 (5/5) A2 (5/5)A4 (5/5) A5 (1/5) A4 (2/5) A6 (2/5) A6 (3/5) A5 (2/5) A7 (5/5) A8 (5/5) A9 (1/5) A1 (1/5) A8 (3/5) A10 (5/5) A z ( x / y ) : x out y individual spectra are identified as sample z 56

PAGE 57

Table 3-3. Detected sim ilarities between samples using p-values obtained by Students T-test (pvalue >0.05) A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 All lines in the spectrum are used, no spectral mask Linear A3 A2 A6 A5 Rank A8 A9 A10 A3 A2 A4 A6 A5 A9 A1 A8 Selected lines in the spectrum are used, spectral mask is applied Linear A3 A2 A6 A5 A9 A8 Rank A8 A9 A3 A2 A4 A6 A5 A9 A1 A8 57

PAGE 58

CHAP TER 4 COMPARATIVE STUDY OF FOUR LIBS SYSTEMS FOR THE ANALYSIS AND IDENTIFICATION OF SOLIDS Introduction Identification and discrimination of individual specimens out of a set of similar samples, can be considered as an important task for forens ic applications and for the industrial sorting of materials. Consequently, as much spectral information as available should be used for an accurate sample identification analysis. This m eans that a spectrometer should be capable of first, resolving individual em ission lines while avoiding spectr al interferences and second, recording a large spectral interval for obtaining a detailed spectral fingerprint of the material of interest. Material identificat ion using LIBS spectra has been demonstrated to successfully discriminate between compositionally close materials [155, 156, 158, 169-176]. Our group has contributed to the development of a LIBS-based method for identification of different types of materials [155-158, 167, 177] The growing interest on LIBS, together with the broad availability of less expensive and more robust lasers, high-resolution spectrometers, and the ease of data collection and processing make this technique very popular in many analy tical labs. Moreover, th ere are already several LIBS systems with the above configuration ava ilable on the commercial ma rket. The final choice of a system is merely determined by the de sirable mode of instru ment operation (laser, spectrometer and detector), the type of application a nd monetary constraints. The purpose of this study is to evaluate th e performance of four spectrometers for the specific task of quantitative compositional analys is and identification of alloys. Three echelle grating spectrometers: Aryelle (LTB Lasertechni k Berlin, Germany), SE200 (Catalina Scientific Instruments, USA) and Mechelle 5000 (Andor Technology, USA), and a conventional broadband grating spectrometer LIBS 2000+ (Ocean Optics, USA) were used sequentially to collect spectra 58

PAGE 59

from the same samples under the same laboratory conditions. Instrumental characteristics such as resolution, LOD, and reproducibility were measur ed and compared. For material identification, correlation analysis was performed using spectra and spectral libraries collected with the same or other spectrometer from the group. The effect of spectrometer characteristics on the robustness of identification was investigated. The results of this study may assist one in choosing an adequate spectroscopic system based on a r easonable compromise between the system performance and its cost. Experimental To fairly compare the performance of the four spectrometers for LIBS analysis, the experiment was carried out usi ng the same laser source and light collection geometry. Due to manufacturers configurations each spectrome ter was coupled to a different detector. LIBS Setup The laser used for these studies was a 1064 nm Q-switched Nd :YAG, model Ultra manufactured by Big Sky Laser Technologies, USA. This laser delivered maximum pulse energy of 50 mJ with a length of 10 ns. The laser beam was focused onto the sample surface using a 5 cm focal length lens, which resulted in an irradiance of ~5.2 GW/cm2 on the sample surface. The radiation emitted by atomic and ionic species abla ted from the sample was collected by a quartz lens and guided into the spectrometers. Figure 4-1 shows the general e xperimental setup used with the four systems. A sample was placed on an XYZ translation st age for sample height adjustment and fresh spot ablation. A laser positioning system consis ting of a diode laser and a photodiode detector ensured a reproducible height positioning of the sample. For data acquisition, each sample was ablated at 10 positions; each position consisted of 35 ablation shots taken at 50 mJ, 10 Hz and atmospheric pressure. The first 10 spectra were 59

PAGE 60

discarded and the next 25 were averaged providi ng an individual spectrum per position on the sam ple. For the correlation studies, 10 individual spectra per sample were averaged to create a sample library. The principal technical characte ristics of the spectrometers an d detectors used in this study together with the delay/gate combina tions are presented in Table 4-1. Samples Seven cast iron standards (SRM 232-239, Cast Irons C18.8, Czechoslovakian Research Institute) were used. These standards are compos itionally close to each other with small (~ <1%) variation in concentrations of trace components (Cr, Ni, Mn, Co Si, etc). The composition of these standards is presented in Table 4-2. Methods Calibration curves for some selected spectral lines representing different elements present in the samples were recorded using matrix-matched standards. For material identification by linear correla tion methods, a home-made program, described in Chapter 3 was used. Results and Discussion Spectral Resolution High spectral resolution is helpful because it reduces spectral interferences caused by other elements and provides more spectral in formation for correlation analysis. The spectra of solid cast iron samples recorded with the four spectrometers yield a huge wealth of information for atomic and ionic lines of elements present in the samples. First we compared the spectral resolutions of the spectrometers. Figur e 4-2 shows spectral fragments obtained with the four spectrometers for the sample cast iron 236 (91.8% Fe and 1.08% Mn). The experiments show that the spectral resolution of the Aryelle system is approximately three 60

PAGE 61

tim es higher. Thus, the Aryelle spectrometer can eliminate spectral interferences such as the FeMn line pair at ~294.8 nm shown in Figure 4-2. This figure also shows th e Aryelle detection of the atomic Fe line at 294.1nm. Although less th an the Aryelle, the Me chelle 5000 spectrometer still shows better spectral reso lution than the SE200 and LIBS2000+ systems. The latter has an approximately 30% lower number of pixels per line compared to the other three systems that, in addition, affects the obtaina ble spectral information. Limits of Detection and Precision Quantitative measurements in LIBS rely on a linear relation between the elemental concentration and the correspond ing emission signal. Calibrati on curves were constructed relating the specific line integrated intensity to the corresponding element concentration present in the cast iron standards. Both, atomic and ionic emission lines were measured for Mn (900 18600 ppm). Atomic lines were studied for Cr (180-19200 ppm), and Cu (380-9200 ppm). Each point of the calibration curve corresponds to the average integral intensity of a spectral line based on 10 subsequent measurements. The inte gral intensity and background s ubtraction were systematically calculated using self-written software in MATLA B. The calibration plots (Figure 4-3) show good linearity within the concentr ation range and high correlation coefficients (R >0.95). The LOD in LIBS are typically poorer than those obtained by other elemental analysis techniques.[178] In the present study the LOD (Table 4-3) were calculated based on the three time standard deviation of the corresponding bla nk measurement. It shoul d be kept in mind, however, that these results are not general but specific to th e experimental equipment and conditions adopted during our experiments. Optim ization of LOD was not an objective in this study; LODs better than those in Table 4-3 have been previously reported using lasers with higher energies [178-181]. 61

PAGE 62

Under these conditions, the LOD obtained with the Aryelle for m anganese and chromium lines were significantly better than those obtaine d with the other three systems. For the copper line studied, the Mechelle 5000 provided the lowest LOD. In general lower LOD were obtained when working with Aryelle, followed by Mechelle 5000, LIBS 2000+ and SE200 in the corresponding order. The missing LOD values in Table 4-3 correspond to lines that were not detected by the specified instruments. Table 4-3 also presents the precision (in parentheses) as the percent relative standard deviation (% RSD) for 10 consecutive measurem ents in sample cast iron 236. There were no mayor differences in precision between the spectrometers. However, lowest % RSD values were obtained by the Aryelle for manganese and ch romium. For the copper line, Mechelle 5000 provides the best precision in the experiments. The LOD obtained were in close agreement to the precision, e.g., lines with larger LOD have larger % RSD values. Material Identification by Linear Correlation This section focuses on the performance of the four spectrometers used for the identification of seven cast iron standards by li near correlation. Ten i ndividual spectra were collected for each sample. Each individual spect rum is the average spectrum from 25 ablation shots. For the correlation analysis, we compare a set of individual spectra vs. a spectral library. We refer to library as the collection of the samp les averaged spectra; hence, a library was created containing 7 averaged spectra. Each of the 70 individual spectra was correlated against the entire spectral library and linear correlation coeffici ents were calculated. The highest correlation coefficient indicated a similarity of a tested spectrum to one from the library. The difference between this and other correlati on coefficients indicates spectral and, hence, compositional differences. 62

PAGE 63

Spectra were correlated us ing only the wavelength range 225-463 nm which was common for all four spectrometers. Excellent identifi cation was achieved for all spectrometers (~100%) when individual spectra were correlated vs. spectral libraries obtained with the same spectrometer (Table 4-4). The somewh at lower performance of LIBS 2000+ (~98%) might be caused by its lower spectral resolution a nd smaller number of pixels per line. Correlation using spectra and spec tral libraries collected with other spectrometers from the group was also studied, e.g. spectra obtained with one spectrometer were compared to the libraries obtained with the other spectrometers. For this purpose, each spectrum, individuals and from the libraries were brought to the same wavelength scale using linear interpolation. Therefore, the converted spectra have the same number of pixels for the group of spectrometers. The use of this type of inte r-correlation didnt pr ovide good identification probably due to the differences in resolution and sensitivity betw een the four spectrometers (Table 4-5). Conclusions In summary, we have carried out prelimin ary studies of the performance comparison between four spectrometers in te rms of spectrochemical figures of merit and their capability for material identification by statis tical correlation methods. We note th at this is not an absolute comparison of systems since each spectrometer was coupled to a different detector which affects the overall system performance. All four spectrometers have essentially the same performance with respect precision. The higher spectral resolution and be tter LOD of the Aryelle system make it more useful for quantitative analysis and material identification especially when samples are of very similar composition. However, the price of this sy stem is higher than the ones of the other spectrometers. Therefore, the fina l choice of LIBS system must also be based on the individual application and budgetary constrains. 63

PAGE 64

Nd:YAG 1064 nm 50mJ 10ns 10HzLaserFocusing lens Iron sample Laser (Ultra, Big Ski Tech.) or SE200 (Catalina Scientific Instruments) LIBS 2000+ (Ocean Optics) Mechelle 5000 (Andor) Aryelle (LTB) Figure 4-1. Experimental setup used to compare the four systems. 64

PAGE 65

C ast iron 236 Aryelle Mechelle5000 SE200 LIBS2000293.8294.0294.2294.4294.6294.8295.0 0.0 0.1 0.2 0.3 Normalized intensityWavelength (nm) MnII 293.9 MnII 294.92 Fe II 294.4 Fe I 294.8 Fe I 294.1 Figure 4-2. Comparison of spectral reso lution for the four spectrometers. 65

PAGE 66

0.00.20.40.60.81.01.21.41.61.82.02.2 0 10 20 30 40 50 60 70 80 90 100 Line integral intensityMn concentration (%) Aryelle R = 0.990.00.20.40.60.81.01.21.41.61.82.02.2 0 20000 40000 60000 80000 100000 120000 140000 Line integral intensityMn concentration (%) Mechelle R = 0.990.00.20.40.60.81.01.21.41.61.82.02.2 0 20000 40000 60000 80000 100000 Line integral intensityMn concentration (%) SE200 R = 0.990.00.20.40.60.81.01.21.41.61.82.02.2 0 4 8 12 16 20 24 28 32 Line integral intensityMn concentration (%) LIBS 2000+R = 0.99 Figure 4-3. Calibration plots for the Mn II ion line at 293.9 nm. 66

PAGE 67

Table 4-1. T echnical characteristics of the LIBS spectrometers used in this study. Characteristic ARYELLE LTB MECHELLE5000 Andor SE200 Catalina Sci. Ins LIBS 2000+ Ocean Optics Grating Echelle Echelle Echelle Conventional Focal Length (mm) 400 195 200 101 Aperture f/10 f/7 f/10 f/4 Entrance (slit width, um) 50 50 25 10 Resolving power ( / ) 12000 4000 3000 4000* Dimensions (mm) 438 x 200 x 232 571 x 165 x 160 127 x 152 x 318 130 x 483 x 350 Weight (kg) 12 12 13 8 Detector CCD (Andor) ICCD (Andor) I CCD (Princeton) CCDs (Ocean Optics) 2048 x 512 pixels 1024 x 1024 pixels 576 x 384 pixels 7 channels x 2048 pixels Delay time ( s) 10** 1 1 1 Gate Non-gated 2 s 2 s Non-gated, 2ms*** Wavelength range 227 465**** 200 975 190-1100 200-980 Price (US $) 70K 50K 50K 30K Wavelength calibration Built-in Hg and Pt spectral lamps Deuterium/Halogen calibration, Hg-Ar light source Deuterium/ Halogen calibration, Hg-Ar light source Deuterium/ Halogen calibration, Hg-Ar light source Calculated @ 400 nm, spect ral resolution = 0.1 nm ** Delay time set by a mechanical chopper non-gated CCD detector. *** Integration time **** This range refers to the instrument ava ilable in our laboratory; other ranges can be covered upon request. 67

PAGE 68

Table 4-2. E lemental composition in percen tages (%) for Cast Iron Standards C18.8, Czechoslovakian Rese arch Institute. Standard C Cr Cu Fe Mg Ni P Si Mn 232 1.93 1.19 0.038 91.859 0.007 0.026 0.009 3.5 0.09 233 2.12 1.92 0.11 92.184 0.005 0.062 0.033 2.59 0.26 234 2.46 0.46 0.275 91.969 0.009 0.305 0.38 2.02 1.39 235 2.73 0.41 0.157 91.772 0.005 0.195 0.78 0.92 1.86 236 2.85 0.05 0.215 91.814 0.075 1.77 0.084 1.65 1.08 237 3.03 0.15 0.545 92.135 0.017 0.7 0.175 1.2 0.13 238 3.36 0.018 0.92 91.897 0.046 1.11 0.052 1.55 0.48 239 4.15 0.052 0.085 91.877 0.038 2.42 0.024 0.27 0.76 Table 4-3. Limits of detection (ppm) and precisio n (% RSD, shown in parentheses) obtained with cast iron standards. Line ARYELLE MECHELLE SE200 LIBS 2000+ Mn II 257.6 76 (3.8) 630 (6.2) 340 (5.6) Mn II 293.9 190 (4.5) 310 (8.1) 360 (8.6) 270 (10) Mn II 294.9 130 (4.3) 250 (8.5) 400 (20) 180 (7.0) Mn I 403.3 260 (6.4) 1600 (12) Cr I 425.4 360 (7.4) 1040 (9.8) 6900 (10) 4200 (7.1) Cr I 427.5 580 (8.9) 2300 (16) 2300 (13) 5700 (7.3) Cu I 324.8 68 (6.9) 39 (5.3) 370 (11) 250 (10) Calibration plots were not constructed for this p eak since they were not detected in the spectra. Table 4-4. Correct Material Iden tification (%) using linear correl ation. Individual spectra were correlated against libraries obtained with the same spectrometer. SPECTROMETER CORRECT IDENTIFICATION (%) ARYELLE 100 MECHELLE 100 SE200 100 LIBS 2000+ 99 68

PAGE 69

69 Table 4-5. Correct Material Id entification (%) using linear corr elation. Converted individual spectra were correlated ag ainst converted libraries. LIBRARIES SET OF INDIVIDUALS ARYELLEMECHELLESE200 LIBS 2000+ ARYELLE 100 79 40 39 MECHELLE 43 100 80 66 SE200 31 51 100 29 LIBS 2000+ 29 64 31 100

PAGE 70

CHAP TER 5 CALIBRATION CURVES FOR THE QUANTITATIVE ANALYSIS OF GLASS Introduction Several methods have been proposed to obt ain quantitative information from LIBS measurements. Most LIBS applications focus on the determination of various trace elements in a relatively constant matrix. At present, no genera lly accepted approach exists for the use of LIBS to determine main and trace components in vary ing matrices. The variation in composition of such samples (e.g., soil, sewage sludge, steel, cement and glass) influences the laser-sample interaction process because of variable absorption, reflect ion and thermal conductivities properties of the sample surface[182]. Thus, th e characteristics of the plasma and the determination of the composition of the sample will be affected [2]. One of the aims of this project is to identify matrix effects occurring in the analysis of glass by LIBS and to investigate several different approaches whic h could lead to the development of an analytical procedure which en ables the correction of these matrix effects to achieve accurate quantitative analysis. Such qua ntitative analysis would also improve the discrimination capability of the technique in the case of the glass samples analyzed. As mentioned earlier, quantitation is one of the anal ytical aspects of LIBS where improvements are needed. In a LIBS experiment, the wavelength resolv ed detection of the emission lines provides quantitative information of the elemental com position of the sample. The sensitivity for each element is influenced by plasma parameters, whic h are in turn strongly in fluenced by the sample matrix[138]. Under the assumption of local ther modynamic equilibrium [170], atomic and ionic states will be populated and depopulated predomin antly by collisions with electrons rather than by radiation [183]. In this case, the excitation (Boltzmann) and ionization (Saha) temperatures 70

PAGE 71

coincid e with the electronic temperature, which is the temperature of the Maxwellian distribution of electron velocities[184]. Thus, the plasma electronic excitation temperature, T and the plasma electron number density, ne, derived from the plasma emission, can be used for a meaningful description of the plasma characteristics. Many methods have been described for determining T[1, 2, 4] The most widely used method, based on the measurement of relative line intensities normalized to their spectroscopic parameters, relies on the so-called Boltzmann plot. In its simplest form, T can be calculated from the intensity ratio of two emission lines[2], orig inating in different upper levels of the same element and ionization stage mnmij ijimn miAgI AgI k EE T ln (5.1) where Ei and Em are the excitation energies (eV) for the upper levels i and m k is the Boltzmann constant (J.K-1), Imn and Iij are the integrated line intensities (e.g.,W.m-3) corresponding to the transitions between the upper levels i or m and the lower levels j, or m respectively; g is the statistical weight, and A (s-1) is the transition probability. By extending the above to several transitions and by linearization of the expressi on, the Boltzmann plot equation is obtained, kT E TU n Ag Ii s s iji ij )( ln ln (5.2) where is the total number density (m-3) of the species s in the plasma and Us(T) is the internal partition function of the species at temperature T. This equation allows th e evaluation of T from the measured intensities of a series of lines, provided that the tran sition probabilities and statistical weights are kn own. By plotting ln ( Iij/ giAij) vs. Ei, T can be obtained from the slope sn 71

PAGE 72

-1/ kT In this case, a goo d correlation of the experime ntal data to a linear fitting is an indication of the validity of the Boltzmann equation[2]. Calibration plots are useful for determining the concentration of species in an unknown sample by comparing the sample to a set of stan dard samples of known concentration. They are constructed by measuring the LIBS intensities in relation to known calibration standards. This approach is the most practical for extracting quantitative information on sample composition, despite the fact that the laser material intera ction is highly matrix dependent and therefore variations in the matrix between the unknown sample and the standard must be minimal. Besides matrix dependence, calibration plots are also susceptible to fluctuations in laser fluence and sample inhomogeneity. As a result, unusual beha viors such as negative slopes and widely scattered experimental points can be observed. [165] The measured integral line intensity (counts), ijI, can be expressed by [2] (T)U eg AFCIS KT E i ij s iji (5.3) where Cs is the concentration of the emitting species in the sample and F is an experimental parameter which takes into account the optical effi ciency of the collection system as well as the plasma density and volume. ijI is proportional to the populat ion of the corresponding energy levels via the transition probability Aij between the upper j and lower i levels of the transition, and the other terms have already been defined (see Eq. 5.2). Eq. (5.3) illustrates that th e concentration values for each element can be obtained by comparing a selected line intensity from an unknown sample to the corresponding one from a certified sample since all factors in the equa tion are common, except for the concentration and the line intensity. 72

PAGE 73

Since seve ral reference glass samples were available, covering a wide range of concentrations, a calibration curve can be constructed which relates the specific line intensity and its corresponding elemental concentration. Using this calibration curve, the line intensity measured for an unknown sample can be re lated to the corres ponding unknown elemental concentration. This equation, how ever, applies to the plasma species concentration. As a result, one must assume that there is a constant rela tionship between elemental concentrations in the sample and species concentrati ons in the plasma among all the standards and unknown materials for this approximation to be satisfactory. If the ma trix of the certified glass samples differs from the unknown samples, the calibration curve approach might fail. Subsequently, other analytical methods are necessary to obtain more accurate results. As mentioned in Chapter 2, the forensic discrimination of glass relies mostly in elemental analysis techniques. Since ther e is a better contro l of the batch components in the glass manufacturing process, the minor and trace el ements present in glass are considered good sources of discrimination between glasses. Valid ated methods for discrimination of glass based on quantitative data exist for ICP-MS and LA-ICP-MS XRF studies of glass are also available in the literature as described in Chapter 2. These methods characterize the unknown glass fragments by providing quantitative information for a menu of elements. The discrimination of glasses is based on the quantitative results. For example, a validated method for LA-ICP-MS made use of an elemental menu of 10 elements: K, Ti, Mn, Rb, Sr, Zr, Ba, La, Ce, and Pb [107]. A recent study by Micro-X-ray fluorescence ( XRF) and LA-ICP-MS for the discrimination of automotive glass focuses on the concentrations of 5 elements: Sr, Zr, Ti, Rb, and Ba [130]. 73

PAGE 74

The applic ation of LIBS to the forensic analysis of glass is relatively new. Some selected published work on the use of LIBS and the discri mination of glass based their analysis in the following: First, identification by using 10 line intensit y ratios from 18 ionic and atomic emission lines from the elements Al, Ba, Ca, Cr, Fe, Mg, Na, Sn, Si and, Ti in automobile float glass. It was possible to discriminate 83% of the glass samples with 99% confidence based on LIBS spectra alone and 96-99% if the samples were discriminated based on LIBS spectra taken in conjunction with RI data at th e same confidence level [108]. La ter, this study was extended for the discrimination of four common glass types (float, headlamp, brown, and side-mirror glass) [128]. Second, identification by correlation analysis (Chapter 4) [177]. The glass fragments were identified from their unique LIBS spectral finge rprints by using statisti cal correlation methods combined with the use of a spectral mask. A 100 % correct identification was achieved at a 95% confidence level. Third, identification by using different combin ations of 10 line intensity ratios from the elements Al, Na, K, Ca, Fe, and Sr. Excellent discrimination results we re obtained (>99%) [130]. The studies mentioned above focused the discri mination into using the spectra as a whole or in the calculation of intensity ratios for some particular elements (Al, K, Sr, Ba, Ca, Cr, Fe, Mg, Na, Sn, Si and, Ti). These approaches provide qualitative information of the glasses. In this chapter, we examine the LIBS spectra of glass from a different perspective. That is, we are looking at the detailed composition in glass for some of the elements mentioned above by using glass standards. This chapter focuses on the construction of conventio nal calibration graphs made by plotting intensities versus concentration for glass samples from different matrices. 74

PAGE 75

Outlie rs or different slopes in the calibration graphs will indicate the potential presence of a matrix effect. Experimental Glass Standards A total of twelve glass standards were used in this study. They all have a unique aspect and represent different chemical matrices. Magnified pictures of these standards are presented in Figures 5-1 to 5-3. The com position of the nine standard reference materials (SRM) from National Institute of Standard and Technology (NIST) used in this study are listed in Tables 5.1 and 5.2. In Table 5-1, the concentrations of the main and minor elements present in different types of glasses: soda-lime, soft-borosilicate and multi-element are given in percent by weight (%) in the solid form. Pictures of these standards are presented in Figure 5-1; soda-lime glasses are transparent while the others (S RM1411 and SRM 1412) are opaque. Glass made from silica sand will normally have a green tone depending on the amount of iron oxide. To decolorize the gl ass, manufacturers add small am ounts of other colorants which produce a complimentary color to green so that the finished product appear colorless.[52] Ancient glass makers used antimony to produce a white opacity in glass [185]. Nowadays, opacity is achieved when SnO, TiO2 or ZrO2 are added to the glass mixture as opacity agents [186]. Other less common opacifiers are Al2O3 [187] and ZnO [188]. A reported combination for opacity is based on the system SiO2-Al2O3-B2O3-CaO-K2O-TiO2 [189]. Both SRM1411 and SRM1412 have no reported values of the known opacity agents, SnO and ZrO2; while SRM 1411 has only a relatively low (120 ppm) concentra tion of Ti. However, they have a higher concentration (Table 5-1) of Al2O3, ZnO, BaO, K2O and B2O3. The presence of Al2O3 (6 7 %) and ZnO (~ 4%) might explain their opacity. 75

PAGE 76

The com position of trace elements present in glass (soda-lime type) is presented in Table 5-2 in g/g (ppm). The nominal glass composition of these glasses (SRMs 610 through 617) is 72% SiO2, 12% CaO, 14% Na2O, and 2% Al2O3. In addition to the elements presented in Table 5-2, these glasses contain the following 25 elemen ts: As, Be, Bi, Cs, Cl, F, Ge, Hf, Hg, Li, Lu, Mg, Nb, P, Pr, Se, S, Te, Tb, Tm, Sn, W, V, Y, and Zr. The SRM names for this group, e.g. SRM 610-611 are only related to the wafer thickness (in mm) which are 3 and 1 mm respectively. Standards of at least 3 mm thickness were used in this study; therefore onl y the first part of the name is employed when referring to them. These glasses are all transparent and colorless but SRM 610 which has a dark blue appearance (Figur e 5-2). This darkness can be explained by the relatively higher concentrati ons (~ 400 ppm) of Co and Fe present in SRM 610 [52]. In addition to the SRMs, three standards design ed to be representative of ancient glass were used in this study. These standards we re provided by Corning Inc. Their chemical composition (%) is presented in Table 5.3. Thes e three standards are da rk and non-transparent; their picture is presented in Figure 5-3. Instrumentation The schematic diagram of the experimental sy stem used for LIBS experiments performed under atmospheric conditions is depicted in Fig. 54. It consisted of a laser, a spectrometer, and intensified charge-couple device (ICCD), detector gating and control electr onics, and a computer for control and data acquisition. A Q-switched Nd:YAG laser (Quantel Brilliant T27; Big Sky Laser Technologies, Inc) operating at 1064 nm was used in this study. Th is laser delivered a maximum of 200 mJ in approximately 5.3 ns at a maximum repetition frequency of 10 Hz. The laser beam is focused onto the sample surface using a 3.5 cm diameter, 10 cm focal length lens. The firing of the laser flash lamp was triggered by a boxcar (SR250; St anford Research Systems). The Q-switch output 76

PAGE 77

served as th e trigger for th e pulse generator opening the shutter and beginning the gated detection. For these experiments, a pulse energy of 90 mJ was used providing an irradiance of ~5 GW/cm2 on the sample surface. The repetition rate was set to 1 Hz. The radiation emitted by atoms ablated from the glass samples is collected by a 5 cm diameter quartz lens with a focal length of 7.5 cm. An adjustable iris was attached to the lens in order to match the F-number of the spectromete r (F/6.5). The plasma emission was then guided to a 0.5 m focal length Czerny-Turner spectrometer (SpectraPro-500i; Acton Research Corp) by using a 35 m entrance slit. The optimization of the micrometer slit dial and slit width values resulted from a previous study carried out for the analysis of differe nt types of samples: aluminum alloys, brass, soil, and powder alloys [190]. The spectrometer is equipped with a 2400 grooves nm-1 grating. The detector is a twodimensional ICCD (576/RB-E; Princeton Instruments) with 576 x 384 pixels. The ICCD controller (ST-138; Princeton Instruments) was used to gate the detection. Data acquisition was controlled using Winspec 32 software (Version 2. 5.18.2; Princeton Instruments) installed on an Intel Pentium 4, 1.80 GHz computer. The instrumental parameters used are: laser power 90 mJ/pulse, the detector delay time and gate width values were selected accordi ng to the emission lines to be analyzed. Results and Discussion Optimization of Experimental Conditions Experimental conditions capable of providi ng high precision and repeatability between experiments are needed to obtain ac curate quantitative information. First, the effect of the laser power is studied. The Nd: YAG laser used has specified maximum pulse energy of 200 mJ. In our experime nts, it is found that th e laser power (90 mJ) resulted in better signal-to-noise ra tios without detector saturation. 77

PAGE 78

In LIBS, small unpredictable experim ental fl uctuations can cause a significant change in appearance of the spectra. To check the stab ility of our measurement set-up the following protocol is envisaged: the plasma image at ze ro-order for SRM 612 is monitored at regular time intervals. If the difference betw een the plasma intensities is be tween the expected statistical variation of our method ( = 15%), there is trust is the measured intensities and the experiments are continued. Fluctuations in the intensities of the plasma image were always below 10% and there was no need of optimizatio n, unless the collection lens was accidentally moved requiring then the entire collection optic s to be re-aligned until the results were satisfactory. The short term pul se-to-pulse energy fluctuation of the laser was not expected to affect the results. The position in the sample, where the plasma is formed, relative to the collection optics and spectrometer entrance slit was shown to affect the signal in this c onfiguration. Since the ablation is perpendicular to the collection optics, this particular configuration is prone to be affected by the geometry of the sample. When ablating in the extreme opposite side of the sample, the signal was of different magnitude th an when ablating at the edge closer to the entrance slit. Figure 5-5 shows the zero-order plasma image for an iron sample (Cast-Iron 236) with 2 cm thickness and 3.5 cm diameter. In a similar way, Figure 5-6 shows the zero-order plasma image for glass (SRM 1831) with a 0.25 cm thickness and 4 cm length. These two images were taken when working in image-m ode for the ICCD camera. Then, there is a distortion of the plasma image when working far from the spectrometer slit. This distortion affects the signal intensity, up to a 25% increase, when working far from the edge and using the entire CCD frame. For small samples (~ 1 cm or less in diameter or length) this distortion was 78

PAGE 79

not observed. The sam ple thickness doesnt seem to contribute to the distortion in the image. Therefore, it is recommended to ablate close to the edge in relatively long samples (> 1cm). Data collection was mainly performed in s pectroscopy mode rather than in image mode for quantitative analysis. Image mode was ma inly used to determine the stability of our measurement set-up and to reproduce the distance from the focusing lens for each sample; this was done by centering the zero-order plasma imag e to the center of the ICCD camera. All plasma images in this study were taken at a dela y time of 2 s and 0.25 s integration time; to prevent ICCD spot damage from intense plasma light emission, a neutral density filter was placed in front of the spectrometer input slit. Fr om this check, it was observed that each standard has a unique plasma image intensity when working in the zero-order grating. Figure 5-7 presents the plasma zero-order image for the twelve glass standards used in this study. Another parameter of interest in this study wa s the number of CCD vertical pixels to be binned during spectra acquisition. Binning is th e process of combining charge from adjacent pixels in a CCD into one large pixel[191]. A sp ectral line is typically an image of the slit formed on the CCD. The signal from a single spectra l line can be binned to achieve the best SNR [190, 192]. The zero-order plasma im age for Corning, Glass C, taken at a delay time of 2 s and 0.25 s integration time is presente d in Figure 5-8. Five groups of pixels have been selected for spectral binning. The effect of binning was determ ined using a Mg I line at 518.4 nm shown in the first panel in Figure 5-9. The signal, b ackground, %RSD, noise and signal-to-noise (S/N) were calculated as a function of the binning grou p and shown in Figure 5-9. Each point is the average of 5 successive experiments performed at different positions in Corning glass C. The Mg peak intensity and background around the peak decrea se as less pixels are summed, behavior that is expected. The corresponding %R SD is higher when the total number of pixels is binned, 79

PAGE 80

having a relative constant value (4-6%) for th e other groups. The noise, calculated as the standard deviation of a background pixel located at the left of th e Mg peak, decreases with the num ber of binned pixels. Since the noise is re duced by binning, the signal-to-noise ratio is improved when fewer pixels are used. From these results, it would make sense that the limit to the number of binned pixels for all succeeding meas urements should be the height of the plasma (100 pixels in this case). However, as Figure 57 shows, the height and intensity of the plasma varies with the type of sample. To ensure the whole plasma is binned in all samples, 200 pixels were chosen as the limit for the subsequent expe riments. This binning set also provides relative good precision, low noise, and adequate signal-to-noise values. There is no sample preparation required for elemental analysis by LIBS. However, relatively low spectral intensities were recorded after the first laser pulses which slowly increased reaching later a constant value. This behavior was specially presented in transparent glass standards. Figure 5-10 presen ts plots of the variation of th e net intensity of Sr for SRM 620 and Corning C. The intensity of a strontium atomic emission line at 460.73 nm was monitored vs. the number of laser pulses (up to 100) taken at one spot in the samples. The behavior in SRM 620 was also observed in Chapter 4. For the first few laser pulses, SRM 620 is almost transparent and there is minimal ablation. However, as th e glass interrogation prog resses, defects in the surface were formed making ablati on stronger. The variation of p eak intensities was calculated for each glass standard. The precision (%RSD) vari ability with the number of laser pulses at one sample spot (in groups) is presented in Tabl e 5-4 for glasses SRM 620 and Corning C. %RDS values observed for SRM 620 were common to ot her transparent glasses; the %RSD improves with the number of laser pulses. After 20 laser pu lses, %RSD has a relative constant value close to 6%. For non-transparent samples (e.g. Cornin g C) there is no major variation in precision 80

PAGE 81

(~6%) with the num ber of laser pulses. Consequently, the first 20 ablation pulses were discarded in all samples and considered as preparation pulses. Construction of Calibration Plots As mentioned earlier in this chapter, the iden tification of glass by elemental analysis relies mostly in the study of elements present in minor or trace concentrations. In order to obtain an accurate characterization of gl ass, as many elements as possible should be taken into consideration. From previous LA-ICP-MS and LIBS studies, the elements most commonly used for discrimination analysis are the following: Al, K, Sr, Ba, Ca, Cr, Fe, Mg, Na, Sn, Si and, Ti. This analysis focuses on the study of Sr, Ti, and Mg. The selection of spectral lines is criti cal for the success of a quantitative LIBS measurement. Spectral resolution, sensitivity, and absence of interferences are important factors to be considered when selecting lines. To exte nd the dynamic range of th e calibration plots, one may select different analyte lines to be used alternatively within a certain concentration range. The stronger analyte lines, even when exhibiting self-absorption at highe r concentrations, could be used in the lower concentration range, wh ile the weaker lines may be used for higher concentrations[193]. In addition, time-resolved detection is importa nt to discriminate the atomic emissions against the strong background radi ation. The background radiation is mainly composed of bremsstrahlung from free electrons and recombination emission which start almost simultaneously with the plasma ignition[182]. The detector delay and gate width or integration time were selected depending on the line to be studied. Determination of strontium Two strontium peaks were selected, Sr atomic (I) present at 460.73 nm and Sr ionic (II) present at 407.77 nm. Glass spectra for two glasses SRM 1411 and Corning D (for comparison 81

PAGE 82

purposes) w ere observed in Figure 5-11 (Sr 460.73 nm, 2 s delay time and 10 s gate width) and Figure 5-12 (Sr 407.77 nm, 2.5 s delay time and 0.75 s gate width). The effect of gate delay afte r the laser pulse on the peak intensity was studied for both lines. The integration time or gate width for detection was set to 0.5 s. Figure 5-13 presents the temporal evolution of Sr I at 460.73 nm. Highe r intensities were obtained when working at delays between 2 to 5 s. Relatively good sensitivity was obs erved at gate widths around 10 s. Consequently, a detector delay of 2 s and an integration time of 10 s maximize the signal of this Sr peak. Figure 5-14 presents the temporal evolution of Sr II at 407.77 nm. Higher peak intensities were obtained when work ing at delays between 1.5 and 2.5 s. For Sr II at 407.77 nm, a detector delay of 2.5 s and an integration time of 0.75 s maximize the signal for the studied concentrations. For data acquisition, the first 20 spectra we re discarded and the next 20 were accumulated per position in the sample. Each point in the calib ration plot is the result of three measurements or positions per sample. Calibration plots for th ese lines are presented in Figure 5-15 (Sr I) and Figure 5-16 (Sr II). It is observed in Figure 5-15 that the peak intensity for Sr I at 460.73 nm in SRM 1411 was higher than expected. To confirm this observation, the calibration plot for this peak was repeated at a later day with similar results (Figure 517). A 7% variation betw een both experiments was observed when averaging 20 laser pulses (from 21 to 40). Besides concentration of the analyte, the inte nsity of peaks in LIBS is also affected by many experimental parameters such as interferen ces or spectral overlap, self-absorption, and matrix effects[194]. One obvious reason for the anomaly observed could be the presence of a spectral interference under the Sr line. Table 5-5 contains a list of spectral lines present in the 82

PAGE 83

vicinity of Sr I 460.73 nm[195]. As m ention earlier, the main difference between SRM 1411 and the other standards is a higher concentration of Al2O3, ZnO, BaO, K2O, and B2O3. Within the spectral bandwidth of our apparatus, there are no lines for Al, O, Zn Ba, and B near this Sr peak. Further experiments, for example, using a hi gher resolution spectrometer to check for an asymmetry in the line shape, are needed to confi rm or refute if the higher Sr signal intensity is due to spectral interferences. If the value for Sr I at 460.73 nm in SR M 1411 is withdrawn from the calibration plot, (Figure 5-15 A) a linear plot w ith a high correlation coefficient is obtained (Figure 5-15 B). Even for Corning C, with a Sr concentration (2500 ppm), the correspondence be tween peak intensity and concentration is linear. Therefore, it is not expected that the hi gher intensity in SRM 1411 might be due to self-abs orption at ~ 800 ppm. Another possibility for this unexpected hi gher Sr peak intensity in SRM 1411 is the potential presence of a matrix effect. SRM 1411 is a soft-borosilicate glass while most of the other glass standards used in the calibrat ion plot are of soda-lime composition. It was then decided to use another Sr analytical line. A new calibration plot was constructed for a Sr ionic line present at 407.77 nm (Figure 5-16). Surpri singly, this time the calibration plot does not show any obvious outlier and the intensity of Sr in SRM 1411 fits nicely in the linear regression plot. The detection limit is defined as the concentration that produces a net line intensity equivalent to three times the standard deviation of the background[193]. It can be evaluated with the expression S LODB3 (5.4) 83

PAGE 84

W here B is the standard deviati on of the background, and S is the sensitivity defined as the slope of the calibration plot for the selected line. For each peak, the standard deviation in the background of the Sr peak was calculated. The glass with the lowest Sr c oncentration SRM 616 (~ 40 ppm) wa s selected for this task. The variation of the Sr intensity at 460.73 nm with the number of laser pulses (from 21 to 40) in SRM 616 is shown in Figure 5-18. Since there is no blank for LIBS measurements, the standard deviation around the Sr 460.73 peak wa s calculated for the 20 spectra shown in Figure 5-18. As reported in the recent li terature[196], if the standard de viation for each pixel is plotted vs. wavelength a plot that resembles the studied spectra is obtained (Figure 5-19). The standard deviation for the background region around the Sr peak has an average value of ~ 3200 counts. By using the slope of the calibration plot for th is peak (Figure 5-15 B), the calculated LOD is 21 2.4 ppm. In a similar way, the LOD for Sr at 407.77 nm was calculated. The plot of standard deviation for the region around 407.77 nm is presen ted in Figure 5-20. The standard deviation for the background region around the Sr peak has an average value of ~ 22 000 counts. By using the slope of the calibration plot for this p eak (Figure 5-16), the calculated LOD is 1.7 0.30 ppm. LODs for Sr determination with LIBS, using the 460.73 nm atomic line, are 2 ppm in wastewater [197] and 80 ppm in soils [198]. The i onic line at 407.77 nm has been mostly used in the quantitative analysis of solid waste (13 ppm)[ 48], soils (30 and 42 ppm) [198, 199] the latest when using a portable LIBS instrument, and Sr in a UO2 pellet (130 ppm) [200]. In a recent study of wavelength dependence on the elemental an alysis of glass, the authors were able to provide quantitative information for vehicle glass samples with Sr concentrations higher than 30 84

PAGE 85

ppm [129]. Sr II at 421.5 nm has also been repo rted in LIBS for the determination of Sr (LOD 0.3 ppm) in starch based flours [200]. Determination of magnesium For the determination of Mg, the atomic line at 517.27 nm was selected (Figure 5-21). A combination of 2 s for both delay and integration time was selected for this line. The calibration plot for peak intensity vs. concentr ation is presented in Figure 5-22. Following the same methodology as with stron tium for LOD determina tion, the standard deviation around the peak was calculated (Figure 5-23) The plot of the standard deviation of the measurements vs. wavelength is presented in Figure 5-24. The calculated LOD was 120 7.5 ppm. This LOD is much lower than those usuall y required for ordinary gl ass analysis, e.g. Mg concentration in vehicle window glasses is a bout 2-3% (3-5% MgO)[99] Reported LIBS LODs for Mg in glass are 130 and 28 ppm in air at 1 and 5 torr respectively when using the peak intensity of Mg I at 383. 83 nm[201]. Most quantitative studies for Mg focus on its determination in aluminum alloys and soils, LODs as low as 0.5 ppm have been reported[193]. Determination of titanium For the determination of Ti, the ionic line at 336.12 nm was selected. Glass spectra for two glasses SRM 1411 and Corning D (for compar ison purposes) are obser ved in Figure 5-25. Optimum delay and integration times were found at 1 and 4 s respectively. The calibration plot for peak intensity vs. concentration is presen ted in Figure 5-26. For LOD calculation, the standard deviation in the background for sta ndard SRM 612 was calcula ted over 20 laser pulses (Fig. 5-27). The plot of standard deviation for 20 laser pulses vs. wavelength (Fig. 5-28) also resembles the shape of the line. The calculated LOD was 19 0.98 ppm. This LOD is much lower than those usually required for ordinary glass analysis, e.g. Ti concentration in vehicle 85

PAGE 86

window glasses is ~ 70 ppm (120 ppm TiO2)[99]. Reported LIBS LODs for Ti in glass are 410 and 350 ppm in air at 1 and 5 torr respectively when using the peak intensity of Ti I at 365.35 nm[201]. Similarly to Mg, most reported st udies of titanium determination are focus on aluminum alloys and soil matrices with reported L ODs as low as 4 ppm in aluminum alloys (Ti II 323.45 nm)[202]. In these experiments, it can be observed that the detection limits are a function of the studies elements and the selected emission line s. This observation can be explained by the following factors[193]: The intensity of the analytical line, which is related to the transition probability. The upper energy of the analytical line (it is more difficult to populate higher energy levels). The spectral surroundings of the analytical line wh ich is related to the detector sensitivity. The detection limits (Table 5-6) for Sr, Mg and Ti were comparable to those reported in LIBS studies in glass, al uminum alloys and soils. Conclusions The feasibility of constructing calibration pl ots for NIST and Corning glass samples was evaluated for Sr, Mg, and Ti. These three elements are common to most types of glasses and its determination can provide useful information for discrimination studies. When determining concentrations, linear calibration curves are desi rable because they result in the best accuracy and precision. All four construc ted calibration plots were linear (R > 0.9) over the studied range of concentrations. Linearity was achieved even when the glass standards were from different matrices: soda-lime, borosilicat e, multi-component and lead-glass. The only exception was the relatively higher intensity of Sr 460.73 nm in SR M 1411. In most cases, re lative errors of less than 10% were obtained. The delay time and gate wi dth chosen for the detector was also an 86

PAGE 87

essential p arameter to conclude on the efficien cy of these measurements. The obtained LODs were significantly lower than thos e required for the analysis of gl ass; therefore, these calibration plots could be use as to determine the quant itative composition of unknown glass fragments 87

PAGE 88

Figure 5-1. SRM-NIST Series 112.3 ( glasses in powder and solid forms) Figure 5-2. SRM-NIST Series 112. 4 (trace elements, wafer form). 88

PAGE 89

Figure 5-3. Corning glass standards. Nd:YAG Laser Pulse Generator and ICCD controller ICCD cameraS p e c t r o m e t e r Translation Stage Air flow controller Figure 5-4. Diagram of the experimental LIBS system used in the experiments. 89

PAGE 90

Cast iron -236 3.5 cm 2.1 cm Figure 5-5. Zero-order plasma images obtained at different positions in standard cast iron 236. SRM 1831 4 cm 0.2 cm Figure 5-6. Zero-order plasma images obtained at different positions in SRM 1831. 90

PAGE 91

Figure 5-7. Zero-order plasma images and their in tensities (a.u) for the glass standards used in this study. 91

PAGE 92

Binning A 386 pixels Binning B 200 pixels Binning C 100 pixels Binning D 40 pixels Binning E 11 pixels Figure 5-8. Zero-order plasma image for Corni ng glass C showing five regions for detector binning. 92

PAGE 93

517.8518.0518.2518.4518.6518.8 2.0x1064.0x1066.0x1068.0x1061.0x1071.2x107 Intensity (cts)Wavelength (nm) Mg I 518.4 nmA-386B-200C-100D-40E-11 0.0 3.0x1066.0x1069.0x1061.2x1071.5x107 Net intensity (cts)Binning set Number of PixelsA-386B-200C-100D-40E-11 0.0 5.0x1051.0x1061.5x1062.0x1062.5x106 Background (cts)Binning set Number of pixels A-386B-200C-100D-40E-11 0 4 8 12 16 20 %RSDBinning set Number of pixelsA-386B-200C-100D-40E-11 0 1x1042x1043x1044x1045x1046x1047x1048x104 NoiseBinning set Number of pixelsA-386B-200C-100D-40E-11 0 200 400 600 800 1000 1200 S/ NBinning set Number of pixels Figure 5-9. Binning effects on the intensity, background, %RSD, noise and signal-to-noise ratio for Mg I line at 518.36 nm. 93

PAGE 94

0102030405060708090100 0 100000 200000 300000 400000 500000 600000 700000 Net intensity (counts)Number of laser pulses SRM 620 0102030405060708090100 0 200000 400000 600000 800000 1000000 1200000 1400000 Net intensity (counts)Number of laser pulses Corning C Figure 5-10. Strontium (460.73 nm) peak intensity dependence with the number of laser pulses for standards SRM 620 and Corning C. 94

PAGE 95

459.5460.0460.5461.0461.5462.0 0.0 2.0x1054.0x1056.0x1058.0x1051.0x1061.2x1061.4x106 Intensity (cts)Wavelength (nm) CorningD SRM1411SrI 460.7 nm Figure 5-11. Strontium peak at 460.73 nm. 406.5407.0407.5408.0408.5409.0 0 1x1072x1073x1074x107 Intensity (cts)Wavelength (nm) CorningD SRM1411SrII 407.8 nm Figure 5-12. Strontium peak at 407.77 nm. 95

PAGE 96

0246810121416 2x1053x1054x1055x1056x1057x1058x105 Net intensity (cts)Delay time (us) Figure 5-13. Optimization of the gate delay for Sr atomic line at 460.73 nm 024681012 0.0 3.0x1066.0x1069.0x1061.2x1071.5x107 Net intensity (cts)Delay time (us) Figure 5-14. Optimization of the gate delay for Sr ionic line at 407.77 nm 96

PAGE 97

05001000150020002500 0.00 2.50x1055.00x1057.50x1051.00x1061.25x1061.50x1061.75x1062.00x106 Net intensity (cts)Sr (ppm) 05001000150020002500 0.00 2.50x1055.00x1057.50x1051.00x1061.25x1061.50x1061.75x1062.00x106 Net intensity (cts)Sr (ppm) 97 05001000150020002500 0.00 2.50x1055.00x1057.50x1051.00x1061.25x1061.50x1061.75x1062.00x106 Net intensity (cts)Sr (ppm)05001000150020002500 0.00 2.50x1055.00x1057.50x1051.00x1061.25x1061.50x1061.75x1062.00x106 Net intensity (cts)Sr (ppm) y = (446.5 22.25) x R = 0.99 A B Figure 5-15. Calibration plot for Sr atomic lin e at 460.73 nm. Linearity is achieved (B) after removing the intensity of SRM 1411 present in A. 97

PAGE 98

05001000150020002500 0.0 2.0x1074.0x1076.0x1078.0x1071.0x108 Net intensity (cts)Sr (ppm)y = (39 002 1024) x R: 0.99 Figure 5-16. Calibration plot for Sr ionic line at 407.77 nm. 98

PAGE 99

2022242628303234363840 5.0x1051.0x1061.5x1062.0x1062.5x1063.0x106 Net intensity (cts)Number of laser pulses per position Average SRM1411data1 SRM1411data2 SRM1411data32022242628303234363840 5.0x1051.0x1061.5x1062.0x1062.5x1063.0x106 Net intensity (cts)Number of laser pulses per position Average SRM1411data1 SRM1411data2 SRM1411data3 Experiment 1 Experiment 2 Figure 5-17. Strontium (460.73 nm) peak intensity variation with the number of laser pulses in SRM 1411. 99

PAGE 100

460.0460.2460.4460.6460.8461.0461.2 8.0x1041.0x1051.2x1051.4x1051.6x105 Peak intensity (cts)Wavelength (nm) Figure 5-18. Strontium (460.73 nm) peak intensity dependence with the number of laser pulses for standards SRM 616. 460.0460.2460.4460.6460.8461.0461.2 2000 4000 6000 8000 10000 12000 14000 Standard deviation (cts) Wavelength (nm) Figure 5-19. Standard deviation vs. wavelength around Sr I 460.73 nm for SRM 616 100

PAGE 101

407.0407.2407.4407.6407.8408.0408.2408.4 0 20000 40000 60000 80000 100000 120000 140000 Standard deviation (cts)Wavelength (nm) Figure 5-20. Standard deviation vs. wave length around Sr II 407.77 nm for SRM 616 516 517 518 519 2.0x1054.0x1056.0x1058.0x1051.0x106 516.763 517.298 518.378Intensity (cts)Wavelength (nm) CorningD SRM1411Mg I Mg I Fe I Figure 5-21. Magnesium peak at 517.27 nm. 101

PAGE 102

0.00.51.01.52.02.53.03.54. 0 1x1052x1053x1054x1055x1056x1057x1058x1050 Net Intensity (cts)Mg %y = (2.19E5 1.4E4) x R = 0.94 Figure 5-22. Calibration plot for Mg atomic line at 517.27 nm. 102

PAGE 103

517.0517.2517.4517.6517.8518.0 60000 80000 100000 120000 140000 517.286Intensity (cts)Wavelength (nm) Figure 5-23. Magnesium peak intensity dependence with the number of la ser pulses for standard SRM 610. 517.0517.2517.4517.6517.8 2000 4000 6000 8000 517.286Standard deviationWavelength (nm) Figure 5-24. Standard deviation vs. wave length for Mg 517.27 nm in SRM 1411. 103

PAGE 104

335.6335.8336.0336.2336.4336.6336.8 5.0x1051.0x1061.5x1062.0x1062.5x1063.0x106 336.117Intensity (cts)Wavelength (nm) CorningD SRM1411 Figure 5-25. Titanium peak at 336.12 nm. 05001000150020002500 0.0 5.0x1051.0x1061.5x1062.0x1062.5x1063.0x106 Net intensity (cts)Ti (ppm)y = (1 272 65) x R = 0.99 Figure 5-26. Calibration plot for Ti ionic line at 336.12 nm. 104

PAGE 105

335.6335.8336.0336.2336.4336.6 100000 120000 140000 160000 180000 200000 220000 336.133Intensity (cts)Wavelength (nm) Figure 5-27. Titanium peak intensity dependence with the number of laser pulses for standard SRM 612. 335.4335.6335.8336.0336.2336.4336.6336.8 5000 10000 15000 20000 25000 30000 35000 40000 Standard deviation (cts)Wavelength (nm) Figure 5-28. Standard deviation vs. wave length for Ti 336.12 nm in SRM 612. 105

PAGE 106

Table 5-1. C hemical composition of the studied soda-lime, soft-borosilicate and multicomponent glass standards from NIST (SRM). Concentration Wt. % SRM620 Soda-Lime Flat SRM1830 Soda-Lime Float SRM1831 Soda-Lime Sheet SRM1411 SoftBorosilicate SRM1412 Multicomponent SiO2 72.1 73.1 73.1 58.0 42.4 PbO 4.40 Al2O3 1.80 0.120 1.21 5.68 7.52 FeO 0.032 0.025 Fe2O3 0.0430 0.121 0.087 0.05 0.031 ZnO 3.85 4.48 CdO 4.38 TiO2 0.018 0.011 0.019 0.02 CaO 7.11 8.56 8.2 2.18 4.53 BaO 5 4.67 Li2O 4.5 MgO 3.69 3.9 3.51 0.33 4.69 K2O 0.41 0.04 0.33 2.97 4.14 Na2O 14.4 13.7 13.3 10.1 4.69 B2O3 10.94 4.53 As2O3 0.056 SO3 0.28 0.26 0.25 SrO 0.09 4.55 106

PAGE 107

Table 5-2. T race elements present in the glass standards NIST (SRM) used in this study. Concentration SRM 610-611 SRM 612-613 SRM 614-615 SRM 616-617 Pb 426 38.57 2.32 1.85 Fe 458 51 *13.3 *11 Zn *433 Cd *0.55 Ti *437 *50.1 *3.1 *2.5 Ba *41 K *461 *64 30 29 B *351 *32 *1.3 *0.2 Sr 515.5 78.4 45.8 41.72 Cu *444 *37.7 1.37 *0.80 Co *390 *35.5 *0.73 Au *25 *5 *0.5 *0.18 Mn 485 *39.6 Ni 458.7 38.8 *0.95 Ru 425.7 *31.4 0.855 *0.100 Ag *254 22 0.42 U 461.5 37.38 0.823 0.0721 Ce *39 Sb *1.06 *0.078 Unless stated otherwise, values are in parts-per-million (ppm). not certified values, information only 107

PAGE 108

Table 5-3. C hemical composition in glass standard s designed to be representative of ancient glass (Corning Inc). Concentration Wt. % Brill B Brill C Brill D SiO2 62.0 34.3 53.9 PbO 0.61 36.7 0.48 Al2O3 4.36 0.87 5.3 Fe2O3 0.34 0.34 0.52 ZnO 0.19 0.052 0.1 TiO2 0.089 0.79 0.38 CaO 8.56 5.07 14.8 BaO 0.12 11.4 0.51 MgO 1.03 2.76 3.94 K2O 1 2.84 11.3 Na2O 17 1.07 1.2 P2O5 0.82 0.14 3.93 SrO 0.019 0.29 0.057 Cu 2.66 1.13 0.38 CoO 0.046 0.18 0.023 MnO 0.25 0.55 NiO 0.099 108

PAGE 109

Table 5-4. P ercentage relative standard devia tion (%RSD) variability with the number of laser pulses on the same spot Number of laser pulses SRM 620 Corning B 1-10 77 6.4 1-20 47 6.3 1-30 39 5.9 1-40 34 6.0 1-50 34 5.9 1-60 28 6.3 1-70 26 6.3 1-80 25 6.3 1-90 23 6.2 1-100 22 6.2 20-30 6.3 7.2 20-40 6.0 6.7 20-50 6.0 6.6 20-60 6.6 6.6 20-70 6.5 6.4 20-80 6.3 6.3 20-90 6.1 6.3 20-100 5.8 6.3 30 40 2.7 6.4 3050 5.4 6.4 30 -60 5.4 6.4 30 70 5.7 6.2 30 -80 5.5 6.2 30-90 5.3 6.1 30-100 5.0 6.2 40 50 7.2 4.8 40 60 6.3 6.0 40 70 6.4 5.6 40 80 6.0 5.7 40 90 5.7 5.9 40 100 5.3 6.0 50 60 3.8 6.6 50 70 6.3 6.0 50 80 5.7 6.0 50 90 5.3 6.0 50 100 4.8 6.1 109

PAGE 110

Table 5-5. L ist of spectral lines present in the vicinity of Sr I 460.73 nm [195]. Line Wavelength (nm) Ni I 460.50 Mn I 460.54 Lu I 460.54 Ni I 460.62 Cr I 460.64 Ce II 460.64 Sm II 460.65 Er I 460.66 Ne II 460.67 Nb I 460.68 Sc I 460.69 W I 460.70 Fe I 460.71 N II 460.72 Sr I 460.73 Li II 460.73 Au I 460.75 W I 460.76 Rh I 460.81 K II 460.85 Kr II 460.85 W I 460.88 Ti II 460.93 Ne I 460.94 Ti I 460.94 Table 5-6. Detection limits for various el ements present in the studied glasses. Element Concentration (ppm) Sr I 460.73 nm 21 2.4 Sr II 407.77 nm 1.7 0.30 Mg I 517.27 nm 120 7.5 Ti II 336.12 nm 19 0.98 110

PAGE 111

CHAP TER 6 NORMALIZATION OF THE SIGNAL FOR QUANTITATIVE LIBS ANALYSIS Introduction Key features which make LIBS an attractive analytical technique include minimal sample preparation, in situ rapid and real time analysis. Good de tection limits and wide dynamic ranges have been demonstrated in a variety of sample ma trices. However, the signal in LIBS is affected by a high continuum background which demands tim e resolved detection, and relatively poor precision caused by the very st rong non-linear nature of the laser-material inte raction [203]. Calibration can be a difficult task for many matrices. The problem of matrix effects has been a ddressed in many publications. For example, Eppler et al.[204] reported the effects of an alyte speciation and matrix composition on the determination of Pb and Ba in soil and sand with the use of LIBS. The form of the chemical compound (carbonate, oxide, sulfate, chloride, or nitrate) and, the bulk sample composition (different proportions of soil and sand in a soil/ sand mixture) were found as factors to influence the emissions from Pb and Ba in both matrices. In their study, the effects of speciation where not correlated to any physical property or with the absorptivity of the compounds. However, changes in the bulk sample composition were related to changes in the concentration of ionized species through perturbation of the electron density. Alt hough some suggestions were made to correct for the matrix effect, no universal solu tion to the problem was proposed[204]. Later, a normalization procedure was presente d by Chaleard et al.[205]. In this approach, quantification and correction for matrix effects is done assuming emission lines to be a function of two parameters: the vaporized mass and the pl asma excitation temperature. In their studies, the ablation mass was determined using an acous tic signal, whereas the excitation temperature was measured by the two-line method. The two li ne method consists on the measurement of the 111

PAGE 112

ratio between two lines of a give n elem ent used as a temperature sensor. The two lines chosen originate from two separate energy levels in order to give a ratio that is sensitive to variations in the excitation temperature. It was demonstrated that normalization of the net emission intensity by both the acoustic signal and the temperature a llowed for a multi-matrix calibration curve with about 5% precision for Cu and Mn in various alloy matrices. Panne et al.[138, 182] proposed the use of te mperature-normalized intensity ratios. This method was proposed for the LIBS analysis of majo r constituents (Si, Al, and Ca) in glass and glass melts. The normalization resulted in lin ear calibration plots a nd improved analytical performance (8% accuracy). Ko et al.[206] examined the possibility of usi ng internal standardiza tion in laser ablationmicrowave-induced plasma (LA-MIP) experiments. In these experiments w ith binary alloys, the authors found that intern al standardization could be successfully applied if the line intensities were measured after a sufficient time delay, allo wing the atomization of the sample material to be completed. This delay time was found to be significantly longer for Cu/Zn alloys (~16 s), where the elements involved have largely differe nt vapor pressures, than, for instance, Fe/Cr alloys. Gornushkin et al. [155, 158] reported on the use of linear a nd rank correlation for reliable material identification. Galbacs et. al. [165] proposed a new calibration approach to analyze binary solid samples at the per centage level. The method is based on the observed dependence of the linear correlation coefficient on the analyte concentration in a binary sample. The linear correlation coefficient is calculated between spec tra of a range of certif ied standards and the spectrum of a reference sample (the analyte in th e form of a pure metal), and the resulting curve is used as a calibration curve. Their method provided 1% accurate results for major components. Later, they report on a matrix-fr ee calibration based on th e normalization of the 112

PAGE 113

em ission intensities by surface density for correc tion of matrix effects in the detection of magnesium in powdered samples[203]. In 1999, Ciucci et al. [207] proposed a new method for standard-less LIBS analysis. The so-called calibration-free laser-induced breakdown spectroscopy (CFLIBS) is as an alternative method to quantitative analysis by conventional calibration curves. This method compensates for matrix effects by applying basic equations derived from the local thermodynamic equilibrium assumption. The method relies on the assumptions of stoichiometric ab lation, local thermal equilibrium and, in its initial formulation, opt ically thin plasmas [2, 207]. Several research groups have applied this and variants of this met hod to the analysis of different types of samples [122, 207-219]. This research aims to a further insight into normalization of the signal for quantitative LIBS analysis; the procedure looks into a possible correlation between the background fluctuation (or plasma continuum) and the analyt ical signal in single-s hot measurements. This method has been proposed by Xu et al. for the det ection of Zn in aerosol samples[220]. The basic idea behind this method is that shot-to-shot signal fluctuations can be described as a multiplicative effect, for both the spectral peaks and the background fluctuation. In the practical implementation of the approach, numerous single-shot spectra were recorded, and spectral line intensities were found to correlate with the c ontinuum plasma background. A simple algebraic model was proposed [220]in which the background ( Bi) for each laser shot i was given as the sum of a constant ( bo) and a fluctuating ( k1fi ) term, where k1 is a proportionality factor and fi is responsible for the signal and baseline fluctuations. The peak height measured for shot i can be expressed as o iikCbkCBP )1 ( (6.1) 113

PAGE 114

where k is a proportionality factor th at correlates the fluctuatio ns with the peak intensity and C is the elem ent concentration. A plot of the peak intensity versus the background baseline intensity should then yield a straight line with a slope ( ) equal to 1+ kC and a plot of ( -1) vs. C would therefore result in a lin ear calibration curve passing th rough the origin and free of fluctuation effects. This plot can be used instead of a traditional calibration curve[220]. Gornushkin et al.[221] applied this method to the analysis of iron, aluminum, phosphorous, silicon in phosphate rock, zinc in brass, and chromium in stainless steel. Although they found some degree of correlation, the correlation coeffici ents were somewhat uncertain due to the large scatter of the experimental points; the plot of ( -1) vs. C was not linear and therefore no improvements were obtained when using this me thodology. Nevertheless, this method seems to be simple to apply and has given good results in the case of aerosol samples[220], and it was therefore considered worth of be ing revisited. Its usefulness in the case of solid samples would be assessed in comparison to conventional calibration plots. Experimental The experimental system and spectral data acquisition used in all measurements have already been described in detail in Chapter 5. The schematic of the setup is illustrated in Fig. 5-4. Aluminum alloys and steel standards were used in these experiments. The aluminum alloy standard disks came from: National Research Council Industrial Mate rials Institute (NRC-IMI) in Canada APEX Smelter Co. in South Africa, and Federal Institute for Materials Research and Testing (BAM) in Germany The elemental compositions for these standards ar e given in tables 6-1 to 6-3, respectively. Besides aluminum alloys, a silic on wafer was used to monitor the relationship between the signal 114

PAGE 115

and the back ground. The composition of the steel st andards (C1-C10) from BAM is presented in Table 6-4. For all the experiments, pulse energy of 90 mJ and a repetition rate of 1 Hz were used. The detector delay time and gate width values were selected according to the emission lines to be analyzed. Results and Discussion To date, most of the research devoted to e nhance LIBS sensitivity and precision is based on averaging of spectra (incl uding data filtering algorithms) as a means to overcome the extensive spectral fluctuations observed on a laser shot-to-shot basis[222]. There are some industrial tasks, e.g. in situ analysis of metallurgic specimen s, requiring a fast, non-contact, and reliable method in where single-shot spectra acquisition by LIBS might become a reliable alternative for analysis[223]. As described in the chapter introduction, if there is an association between the spectral peak and the background fluctuation the slope of th is plot can be used to normalized the signal for single-shot experiments. Severa l experiments were carried out to confirm this association. In the top part of Figure 6-1, a 10 nm spectral wind ow from 246 to 256 nm is presented for steel with an approximate content of 0.5% Si (BAM-C1). In the bottom part, a positive zoom of the Si peak at 251.61 nm is shown. This peak is selected for the subsequent analysis. It is observed in Fig 6-1 that the spectrum is very rich in lines and the background around th e selected Si line is not as uniform as desired. To avoid using the wi ngs of the line itself when calculating the true background, several so-called backgr ounds were tested (Figure 6-2): Vicinity of the line, Bi. Left of the peak, Bleft. For this line, Bleft is of relatively lower intensity than the background at the right (Fig. 6-1). 115

PAGE 116

Spectral window (~ 10 nm), Bi,average. Net background, Bi,net; (Bi,net = Bi Bm) where Bm refers to the minimum number of counts present in the spectral window. Figure 6-3 presents the plots of the peak intensity (Pi) vs. the four backgrounds described above for 50 individual measurements ( 50 spectra) in steel C1 taken at 6 s delay time and 6 s integration time for the detector. Good linear corr elation is observed in all four cases, with correlation coefficients higher than 0.96. In addition, Figure 6-4 shows a good correlation between Bi and Bi,average. Consequently, the back ground around the line, Bi, is used for all succeeding experiments. The fluctuation of the peak intensity for Si at 251.61 nm in steel with the number of individual laser pulses is presente d in Figure 6-5. The calculated %RSD for these 50 spectra is ~ 40%. Figure 6-6 shows the fluctuation of the background, Bi, vs. the number of individual spectra. If these two plots are comp ared against each other (Fig. 6-5 and 6-6), no immediate similarities arise. However, if a consta nt arbitrary value, which can be identified with bo in equation 6-1, is subtracted in Fig. 6-6, a simila r fluctuation trend is observed for both, intensity and background (Fig. 6-7). This information is utilized in the forthcoming data analysis. Calibration plot for silicon in steel For most of the 10 steel st andards, the plot of Pi vs. Bi for Si at 251.61 nm is linear when all 50 single-shot spectra are used (Figure 6-8). There were some cases in which linearity is present for most spectra but for some points of higher background values, the signal doesnt increase accordingly (Figure 6-9) This behavior was only present in some of the standards and for less than five points (spectra) in the plots. Spectra were acquired to build a conventional (similar to Chapter 5 calibration plots) and a normalized calibration plot for Si using the 10 steel standards. For this spectral line, a detector 116

PAGE 117

delay and a gate width of 6 s were chosen. Fifty single-shot sp ectra were acquired p er triplicate per sample. For a conventional calibrati on plot, the average of the net peak intensity (Pi) minus Bi is plotted as a function of the Si % concentration (Figure 6-10). For the normalized calibration, ( 1) is plotted vs. the Si % con centration (Figure 6-11). By compar ing these two plots, the spread of the data is considerably higher for the conve ntional calibration scheme shown in Fig. 6-10. For the normalized approach, one of the points (~1. 5% Si) is an outlier to the linear fit. The reasons for this anomaly are not understood: appa rently, the normalized approach presents a smaller linear dynamic range than the conventiona l approach. However, it is clear that the normalized calibration plot has a higher accuracy than the conventional one. Calibration plots in aluminum alloys This model was also applied to the LIBS anal ysis of aluminum alloys, seven emission lines corresponding to the elements Cr, Mg, Fe, Sn, Mn and Si were used in these experiments (Table 6-5). For each aluminum alloy, 50 single-shot spectra were acqui red per triplicate. For all the lines, a detector delay of 2 s and a gate width of 0.5 s were chosen. Figures 6-12 to 6-18 present a comparison of the pe rformance of the conventional calibration approach vs. the normalized method. For both methods, there is a linear correspondence with the element concentration. Nevertheless, for Si in alumin um (Fig. 6-18), the normalized method is more susceptible to higher Si concentrations and two points are outliers fo r the linear fit. In each figure, the calibration equation: y = ( m m). x together with the correlation coefficient of the linear fit is presented. The correlation coefficient, the value of the slope, m and its associated error, m, are used to compare the accuracy of the two plots for each line. Table 66 is a summary of the results present in Fig 610 to 6-18. For most cases, there are improvements 117

PAGE 118

when using the norm alized approach. The best improvement is for Si in steel (Figs. 6-11 and 612). However for Mg, Fe and Mn in aluminum a lloys, the conventional method performs slightly better. Correlation coefficient as a function of the detector delay time To better understand the conditions for the existence of a corr elation between the background (or plasma continuum) fluctuations a nd those of the analytical signal the following experiment was envisaged. The intensity of a Si peak at 298.76 nm pr esent in silicon wafer (Figure 6-19) was monitored at different dete ctor delay times in the range of 0.5 to 50 s. The gate width of the detector was fixed to 1 s. 200 single-shot spectra we re acquired per delay time. A correlation plot is then obtained for each dela y time (Figure 6-20). As expected, the intensity of the background or continuum decreases with the increas ing delay time; the linear correspondence between the peak in tensity and the background is pr actically lost for delay times higher than 6 s. A plot of the correlation coefficient as a function of the detector delay time is shown in Figure 6-21. The experimental conditio ns used for spectra acquisition affect this normalized approach. There is no correlation be tween signal and background for higher delay times. Conclusions The feasibility of constructing normalized calibration plots base d on the use of the relationship between the peak a nd background intensity was evaluated for Si (in steel) and Cr, Mg, Fe, Sn, Mn and Si in aluminum alloys. The normalized plots were compared to the conventional calibration plots obt ained by averaging the 50 indivi dual spectra. When using the normalized approach, some improvements were observed especially in the case of Si in steel. However, the improvements observed were not systemic and general for all the cases investigated: this seems to indicate that the approach has to be considered case by case. Our 118

PAGE 119

research indicate, however, that the correla tion between the peak signal and the background depends strongly upon the delay tim e chosen fo r the measurement (see Figure 6-21) and upon the availability of spectrally well isolated lines to allow a correct measurement of the background. 119

PAGE 120

246248250252254256 35000 40000 45000 50000 55000 60000 65000 70000 IntensityWavelength (nm) Steel (C1) Si 0.46% 251.50251.55251.60251.65251.70251.75251.80 40000 45000 50000 55000 60000 65000 251.672IntensityWavelength (nm) Si I Figure 6-1. Spectrum of steel in the wavelength range 246-256 nm 120

PAGE 121

251.0251.2251.4251.6251.8252.0 251.141 251.672Intensity (cts)Wavelength (nm) zero counts dark counts Bmin(bo) Pi Bmax Bi Bi, average Bi,net: Bi-bo Figure 6-2. Background(s) for Si I 251.61 nm in steel 121

PAGE 122

0100002000030000400005000060000 0 20000 40000 60000 80000 100000 120000 140000 160000 PiBleftR = 0.96010000200003000040000500006000070000 0 20000 40000 60000 80000 100000 120000 140000 160000 PiBiR = 0.970100002000030000400005000060000 0 20000 40000 60000 80000 100000 120000 140000 160000 PiBi, averageR = 0.9705000100001500020000 0 20000 40000 60000 80000 100000 120000 140000 160000 PiBi, net R = 0.98 Figure 6-3. Linear correlation between Pi for Si at 251.61 nm in steel C1 and the correspondent background. 122

PAGE 123

350004000045000500005500060000 35000 40000 45000 50000 55000 60000 65000 70000 BiBi, averageR = 0.99 Figure 6-4. Linear correlation betwee n background around the Si 251.61 nm (Bi) and in the spectral window from 246-256 nm (Bi,average). 123

PAGE 124

10 20 30 40 50 0 20000 40000 60000 80000 100000 120000 140000 160000 180000 200000 IntensityNumber of individual spectra in steel Si peak Figure 6-5. Fluctuations of inte nsity of Si at 251.61 nm in steel for 50 individual spectra. 10 20 30 40 50 0 10000 20000 30000 40000 50000 60000 70000 80000 90000 100000 IntensityNumber of individual spectra in steel Background intensity Figure 6-6. Fluctuations of background intensity of Si at 251.61 nm in steel for 50 individual spectra. 10 20 30 40 50 0 10000 20000 30000 40000 Intensity Number of individual spectra Background Figure 6-7. Fluctuations of inte nsity of background minus an arbitr ary constant value in steel for 50 individual spectra 124

PAGE 125

40000450005000055000600006500070000 40000 60000 80000 100000 120000 140000 160000 180000 200000 Steel C2 0.37% Si R = 0.98PiBi Figure 6-8. Peak intensity as a function of the ba ckground. Linear correlation is observed for Si at 251.61 nm in steel C2 400005000060000700008000090000100000110000 50000 100000 150000 200000 250000 300000 Steel C6 0.40% Si R = 0.93PiBi Figure 6-9. Peak intensity as a function of the ba ckground. Linear correlati on is not achieved in all spectra for Si at 251.61 nm in steel C6 125

PAGE 126

0.00.20.40.60.81.01.21.41.6 0 1x1062x1063x1064x1065x1066x1067x106 Net intensity (cts)Si concentration (%)y = (4.9E6 7.5E5) x R = 0.7 Figure 6-10. Calibration plot for Si atomic line at 251.61 nm in steel 0.00.20.40.60.81.01.21.41.61.8 0 1 2 3 4 5 6 7 Si concentration (%)y = (8.4 0.4) x R = 0.9 Figure 6-11. Plot of -1 against Si concentration in steel 126

PAGE 127

0 00 10 20 30 4 0.0 5.0x1031.0x1041.5x1042.0x1042.5x1043.0x104 Net intensity (cts)Cr concentration(%)y = (4.7E4 4.0E3) x R = 0.940.0 0.1 0.2 0.3 0.4 0.0 0.4 0.8 1.2 1.6 1Cr concentration (%)y = (3.2 0.2) x R = 0.97 Figure 6-12. Comparison of conventional and normali zed calibration plots for Cr at 301.52 nm in aluminum alloys. 127

PAGE 128

0.0 0.1 0.2 0.3 0.4 0.0 5.0x1041.0x1051.5x1052.0x1052.5x1053.0x1053.5x105 Net intensity (cts)Cr concentration (%)y = (7.0E5 2.9E4) x R = 0.990.0 0.1 0.2 0.3 0.4 0 1 2 3 4 5 6 7 -1Cr concentration (%)y = (16 0.54) x R = 0.99 Figure 6-13. Comparison of conventional and normali zed calibration plots for Cr at 283.56 nm in aluminum alloys. 128

PAGE 129

0.00.20.40.60.81.01.2 0.0 5.0x1051.0x1061.5x1062.0x1062.5x1063.0x106 Net intensity (cts)Mg concentration (%)0.00.20.40.60.81.01.2 0 10 20 30 40 50 -1Mg concentration( %)y = (2.0E6 7.8E4) x R = 0.99 y = (40 2.9) x R = 0.98 Figure 6-14. Comparison of conventional and norm alized calibration plots for Mg at 285.21 nm in aluminum alloys. 129

PAGE 130

01234 0.0 4.0x1048.0x1041.2x1051.6x1052.0x105 Net intensity (cts)Fe concentration (%)0.00.51.01.52.02.53.03.54.0 0 2 4 6 -1Fe concentration (%)y = (4.1E4 2.4E3) x R = 0.97 y = (2.0 0.22) x R = 0.95 Figure 6-15. Comparison of conventional and normali zed calibration plots for Fe at 302.11 nm in aluminum alloys. 130

PAGE 131

0.000.050.100.150.200.250.300.35 0.0 2.0x1044.0x1046.0x1048.0x1041.0x1051.2x1051.4x1051.6x105 Net intensity (cts)Sn concentration (%)y = (3.0E5 4.6E4) x R = 0.860.000.050.100.150.200.250.300.35 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 -1Sn concentration(%)y = (8.7 0.7) x R = 0.93 Figure 6-16. Comparison of conventional and normalized calibration plots for Sn at 283.99 nm in aluminum alloys. 131

PAGE 132

0.00.10.20.30.40.50.60.70.8 0 1 2 3 4 -1Mn concentration( %)0.00.10.20.30.40.50.60.70.8 0.0 2.0x1044.0x1046.0x1048.0x1041.0x1051.2x1051.4x105 Net intensity (cts)Mn concentration (%)y = (1.5E5 6.8E3) x R = 0.98 y = (4.9 0.25) x R = 0.97 Figure 6-17. Comparison of conventional and norm alized calibration plots for Mn at 288.96 nm in aluminum alloys. 132

PAGE 133

02468101214161820 0.0 5.0x1051.0x1061.5x1062.0x1062.5x1063.0x106 Net intensity (cts)Si concentration (%)y = (1.4E5 8.6E3) x R = 0.9602468101214161820 0 10 20 30 40 50 1Si concentration(%)y = (3.9 0.23) x R = 0.99 Figure 6-18. Comparison of conventional and norma lized calibration plots for Si at 288.16 nm in aluminum alloys. 133

PAGE 134

298 299 3005.0x1051.0x1061.5x1062.0x106 298.778Intensity (cts)Wavelength (nm)Average of 200 spectra delay 0.5 s Figure 6-19. Spectrum of Si at 298.76 nm in silicon wafer. 134

PAGE 135

1.6x1051.8x1052.0x1052.2x1052.4x1052.6x1051.2x1061.5x1061.8x1062.1x1062.4x1062.7x106 PiBiDetector delay: 0.5 s R = 0.97130000140000150000160000 1.4x1061.6x1061.8x1062.0x1062.2x1062.4x1062.6x1062.8x106 PiBiDetector delay: 1.0 s R = 0.932.5x1043.0x1043.5x1044.0x1044.5x1041000000 1250000 1500000 1750000 2000000 2250000 PiBiDetector delay: 2.0 s R = 0.9360007000800090001000011000 3x1054x1055x1056x1057x1058x105 PiBiDetector delay: 4.0 s R = 0.91 1800200022002400260028003000 1.6x1051.8x1052.0x1052.2x1052.4x1052.6x1052.8x105 PiBi350040004500500055006000 3.0x1053.5x1054.0x1054.5x1055.0x1055.5x105 PiBiDetector delay: 6.0 s R = 0.76 Detector delay: 10.0 s R = 0.6640060080010001200 2x1033x1034x1035x1036x1037x1038x1039x103 PiBiDetector delay: 50.0 s R = 0.37 Figure 6-20. Peak intensity as a function of th e background for Si at different delay times. 135

PAGE 136

01020304050 0.0 0.2 0.4 0.6 0.8 1.0 Correlation coefficient (R)Delay time (s) Figure 6-21. Correlation coefficient as a function of the delay time. 136

PAGE 137

Table 6-1. C hemical composition in aluminum alloys NRC-IMI Concentration Wt. % N3005 N1075 N4104 MN397 Al 97.6 99.7 87.9 98.8 Si 0.084 0.079 9.67 0.049 Mg 0.37 0.004 1.33 0.012 Cu 0.44 0.007 0.15 0.0008 Zn 0.041 0.014 0.088 0.24 Fe 0.2 0.15 0.6 0.095 Mn 1.13 0.007 0.052 0.6 Ni 0.026 0.006 0.026 0.008 Ti 0.05 0.011 0.024 0.14 Cr 0.026 0.006 0.026 0.008 Sn 0.005 0.003 0.005 0.006 Pb 0.005 0.003 0.005 0.006 Bi 0.006 0.003 0.09 0.007 Zr 0.006 0.004 0.005 0.007 Table 6-2. Chemical composition in aluminum alloys APEX Smelter Co. Concentration Wt. % B8 D33 M7 R14 AA3 D28 SM9 SM10 S11 Al 87.9 84.9 85.47 79.59 69.14 81.55 85.34 84.67 89.20 Si 2.33 8.54 0.52 14 17 9.66 1.69 2.92 0.45 Mg 0.076 0.038 0.06 0.87 0.2 0.004 0.43 1.08 1.11 Cu 6.95 2.89 11.12 2.05 8 1.76 3 2.8 0.98 Zn 0.52 0.59 0.51 0.48 3.2 3.6 3.7 5.45 6.85 Fe 0.8 1.15 1.28 0.63 1.77 0.98 3.7 1.96 0.57 Mn 0.4 0.4 0.34 0.92 0.21 0.59 0.76 0.295 0.5 Ni 0.2 0.5 0.205 0.97 0.106 0.43 0.2 0.065 0.1 Ti 0.16 0.055 0.065 0.16 0.078 0.033 0.07 0.055 0.065 Cr 0.17 0.047 0.05 0.11 0.1 0.21 0.38 0.2 0.115 Sn 0.155 0.048 0.105 0.12 0.12 0.3 0.31 0.26 Pb 0.165 0.14 0.11 0.1 0.08 0.34 0.32 0.245 Bi 0.099 0.68 0.54 Zr 0.17 0.123 137

PAGE 138

Table 6-3. C hemical composition in aluminum alloys BAM Concentration Wt. % 308 309 314 Si 0.0707 11.76 11.49 Mg 2.29 0.00068 0.18 Cu 1.315 0.0048 2.07 Zn 5.67 0.0038 1.19 Fe 0.1634 0.0883 0.76 Mn 0.0342 0.0548 0.4 Ni 0.0122 0.00087 0.22 Ti 0.0285 0.0556 0.16 Cr 0.1962 0.00047 0.05 Sn 0.199 Pb 0.221 Bi 94 g/g Zr 0.0078 55 g/g Table 6-4. Chemical composition in steel standards BAM Concentration Wt. % C Si Mn Cr Ni Mo Co C1 0.092 0.46 0.74 12.35 12.55 --C2 0.0103 0.374 0.686 14.727 6.124 0.0138 -C3 0.0345 0.463 0.722 11.888 12.85 0.0304 -C4 0.019 0.27 1.4 18.46 10.2 0.265 0.116 C5 0.086 0.57 0.791 25.39 20.05 -0.054 C6 0.066 0.405 1.38 17.31 9.24 0.092 0.053 C7 0.0141 0.48 1.311 17.84 10.2 2.776 0.0184 C8 0.143 1.41 1.7 17.96 8.9 -0.018 C9 0.05 0.21 0.89 14.14 5.66 1.59 0.22 C10 0.0201 0.537 1.745 16.811 10.72 2.111 0.0525 Table 6-5. Spectral emission lines for the analysis of aluminum alloys Element Wavelength (nm) Cr I 301.52 Cr II 283.56 Mg I 285.21 Fe I 302.11 Sn I 283.99 Mn II 288.96 Si I 288.16 138

PAGE 139

Table 6-6. R esults for the calibration curves and the analysis of Si, Cr, Mg, Fe, Sn and Mn Element Wavelength (nm) Conventional Normalized Si I in steel 251.61 y = (4.9E6 7.5E5) x %RSD = 15.3 R = 0.72 y = (8.4 0.4) x %RSD = 4.76 R = 0.89 Cr I in aluminum alloys 301.52 y = (4.7E4 4.0E3) x %RSD = 8.51 R = 0.94 y = (3.2 0.2) x %RSD = 6.25 R = 0.97 Cr II in aluminum alloys 283.56 y = (7.0E5 2.9E4) x %RSD = 4.14 R = 0.99 y = (16 0.54) x %RSD = 3.37 R = 0.99 Mg I in aluminum alloys 285.21 y = (2.0E6 7.8E4) x %RSD = 3.92 R = 0.99 y = (40 2.9) x %RSD = 7.25 R = 0.98 Fe I in aluminum alloys 302.11 y = (4.1E4 2.4E3) x %RSD = 5.85 R = 0.97 y = (2.0 0.22) x %RSD = 11.00 R = 0.95 Sn I in aluminum alloys 283.99 y = (3.0E5 4.6E4) x %RSD = 15.3 R = 0.86 y = (8.7 0.7) x %RSD = 8.05 R = 0.93 Mn II in aluminum alloys 288.96 y = (1.5E5 6.8E3) x %RSD = 4.53 R = 0.98 y = (4.9 0.25) x %RSD = 5.10 R = 0.97 Si I in aluminum alloys 288.16 y = (1.4E5 8.6E3) x %RSD = 6.14 R = 0.96 y = (3.9 0.23) x %RSD = 5.90 R = 0.99 139

PAGE 140

CHAP TER 7 CONCLUSIONS AND FUTURE WORK Summary and Concluding Remarks An extensive experimental procedure for the ch aracterization of solid samples, glasses in particular, by LIBS was carried out in this research. There are a number of satisfactory techniques available for the elemental analysis of glass fragments for forensic purposes. In the first part of this is work, linear and rank correlati on techniques were appl ied for discrimination of LIBS spectra from glass samples with similar chemical composition, some of them from the same source. The robustness of this techni que was demonstrated by the 100% correct identification (95% confidence level for a type 1 error) obtained by linear correlation when used in combination with a spectral mask. The iden tification was reliable even when experiments were performed on different days when ambient c onditions might be different and affect the line intensities in the LIBS spectra. The rationale of using spectral masking is to eliminate regions of the spectra containing several intense lines common to all samples and to take advantage of the trace element impurities present in these glasses. We are aware of the fact that there are more sophisticated ways to generate a mask than the one used in this study. Ho wever, it was felt useful to focus at first on a simple masking procedure to see whether any furthe r elaboration of this concept was worth pursuing. More refined procedures are planned for the future. The elemental analysis of glass by LIBS has the potential of becoming a useful technique for the discrimination of forensic glasses. Its usefulness as an analytical method for legal purposes will be determined by its general accep tance in the relevant scientific community. Evidence of general acceptance normally incl udes known error rates and publication of the methods in peer-reviewed journals[168]. These legal aspects and corresponding implications, which would require more in-depth statistical analysis, have not been considered in this work. 140

PAGE 141

In the cou rse of this study, several key instrumental parameters were identified and investigated, and a comparative study regardi ng the performance of up to four different commercial instruments for LIBS analysis of so lids has been made. Results indicate that the spectral resolution and sensitivity of the systems are linked to th e performance and suitability of the technique for material iden tification by correlation methods, especially when samples are of very similar composition. While the qualitative characterization of materials w ithout sample preparation is certainly one of the main advantages of LIBS, the possibi lity of performing accurate quantitative analysis relies on the use of calibration curves made with matrix-matched standards[1, 2]. Such quantitative analysis would also improve the disc rimination capability of the technique in the case of the glass samples analyzed. The feasibilit y of constructing calibration plots for NIST and Corning glass samples was evaluated for Sr, Mg and Ti. These three elements are common to most types of glasses and their determination can provide useful information for discrimination studies. All constructed calibra tion plots were linear (R > 0. 9) over the studied range of concentrations. In most cases, relative errors of less than 10% were obtained. The delay time and gate width chosen for the detector was al so an essential parameter to conclude on the efficiency of these measurements. The obtained LODs were significantly lower than those required for the analysis of glass; therefore, these calibration plots could be used as to determine the quantitative composition of unknown glass fragments. In addition to the conventional analytical procedures followed in LIBS analysis, a normalization procedure of the signa l [220] was investigated in th is research. The viability of constructing normalized calibration plots based on the use of the relations hip between the peak and background intensity was evaluated for Si (i n steel) and Cr, Mg, Fe, Sn, Mn and Si in 141

PAGE 142

alum inum alloys. The normalized plots were compared to the conven tional calibration plots obtained by averaging the 50 individual spec tra. When using the normalized approach, improvements were obtained for most lines especially in the case of Si in steel. The correlation between signal and background decreases with the increasing detector delay time. The results of all these investigations have allo wed a fair assessment of the applicability of LIBS for discrimination and quantit ative characterization of solid samples in general, and glasses in particular. Future Research Directions As an addendum to the research carried out in this dissertation, it is suggested that, for the discrimination analysis, masks for spectral regions instead of lines should be used. It would also be beneficial to focus the analysis to the lines of the minor components common to these glasses instead of blocking the lines of the ones pres ent in higher concentration. To improve the discrimination capability of the technique, the constructed calibration plots can be used to calculate the concentration of unknown glass fragments. The normalization based on the correlation of intensity and background should also be applied to the analysis of gla sses and the results can be compared to the conventional curves. Some preliminary results, not reported in th is Dissertation, suggest the use of another normalization procedure based on the observed beha vior of the signals obtained for selected elements, present in the same concentration (or in a narrow range of concentrations) in different matrices, as a function of the en ergy of the excited state of the transitions investigated. This approach is similar to that used in doubl e-pulse LIBS to interpret the enhancements observed[202], and has not been tested so far with the LIBS technique. Two matrices are used to construct independent calibration curves for each element investigated. For any given element, the plot of the intensity ratio in the two matrices versus the excitation energy will reveal whether 142

PAGE 143

m atrix effects are present as well which physic al parameter (temperature, ablated mass) is involved. Finally, as shown in a recen t publication [196], by analyzi ng many spectra obtained under the same conditions, the noise characteristics of the measurements could be investigated by plotting the standard deviation (S D) and the relative standard deviation (RSD), calculated at every spectral element (pixel) of the selected sp ectral range, as a function of wavelength, for different delay times [196]. This procedure, which would complement ou r correlation analysis, appears to be useful to characterize the limiting type of noise present in our experiments. 143

PAGE 144

LIST OF REFERE NCES [1] Cremers DA, Radziemski LJ. Handbook of laser-induced breakdown spectroscopy. Chichester: John Wiley 2006. [2] Miziolek AW, Palleschi V, Schechter I. Laser-induced breakdow n spectroscopy (LIBS): fundamentals and applications. Cambri dge: Cambridge University Press 2006. [3] Lee Y-I, Song K, Sneddon J. Laser-induced breakdown spectrometry. Huntington: Nova Science Publishers 2000. [4] Singh JP, Thakur SN. Laser-induced brea kdown spectroscopy. Amsterdam: Elsevier 2007. [5] Bogue RW. Boom time for LIBS te chnology. Sensor Review. 2004; 24:353. [6] DeMichelis C. Laser induced gas breakdown: A bibliographical review. IEEE Journal of Quantum Electronics. 1969; 5(4):188-202. [7] Lee WB, Wu JY, Lee YI, Sneddon J. Recen t applications of laser-induced breakdown spectrometry: A review of material appro aches. Appl Spectrosc Rev. 2004; 39(1):27-97. [8] Lee YI, Yoo YJ, Sneddon J. Recent advances in laser-induced breakdown spectrometry. Spectroscopy. 1998; 13(7):14. [9] Majidi V, Joseph MR. Spectroscopic applications of laser-induced plasmas. Crit Rev Anal Chem. 1992; 23(3):143-62. [10] Radziemski LJ. From laser to LIBS, th e path of technology deve lopment. Spectrochim Acta, Part B. 2002; 57(7):1109-13. [11] Rusak DA, Castle BC, Smith BW, Winefordne r JD. Recent trends and the future of laserinduced plasma spectroscopy. TrAC, Tr ends Anal Chem. 1998; 17(8-9):453-61. [12] Scott RH, Strasheim A. Laser induced plas mas for analytical spectroscopy. Spectrochim Acta, Part B. 1970; 25(7):311-16. [13] Sneddon J, Lee Y-I. Laser-induced brea kdown spectrometry. The Chemical Educator. 1998; 3(6):1-7. [14] Song K, Lee Y-I, Sneddon J. Applicati ons of laser-induced breakdown spectrometry. Appl Spectrosc Rev. 1997; 32(3):183 235. [15] Winefordner JD, Gornushkin IB, Correll T, Gibb E, Smith BW, Omenetto N. Comparing several atomic spectrometric methods to the super stars: special emphasis on laser induced breakdown spectrometry, LIBS, a futu re super star. J Anal At Spectrom. 2004; 19(9):1061-83. 144

PAGE 145

[16] Sneddon J, Lee YI. Novel and recent applic ations of elemental determination by laserinduced breakdown spectrometry. Anal Lett. 1999; 32(11):2143-62. [17] Song K, Lee YI, Sneddon J. Recent developm ents in instrumentation for laser induced breakdown spectroscopy. Appl Spectrosc Rev. 2002; 37(1):89-117. [18] Tognoni E, Palleschi V, Corsi M, Cristo foretti G. Quantitative micro-analysis by laserinduced breakdown spectroscopy: a review of the experiment al approaches. Spectrochim Acta, Part B. 2002; 57(7):1115-30. [19] Kaiser J, Galiov M, Novotn K, Cervenka R, Reale L, Novotn J, Liska M, Samek O, Kanick V, Hrdlicka A, Stejskal K, Adam V, Kizek R. Mapping of lead, magnesium and copper accumulation in plan t tissues by laser-induced breakdown spectroscopy and laserablation inductively coupled plasma mass sp ectrometry. Spectrochim Acta, Part B. 2009; 64(1):67-73. [20] Rusak DA, Castle BC, Smith BW, Winefo rdner JD. Fundamentals and applications of laser-induced breakdown spectroscopy. Crit Rev Anal Chem. 1997; 27(4):257 90. [21] Vadillo JM, Laserna JJ. Laser-induced plasma spectrometry: truly a surface analytical tool. Spectrochim Acta, Part B. 2004; 59(2):147-61. [22] Martin MZ, Cheng MD, Martin RC. Ae rosol measurement by laser-induced plasma technique: A review. Aerosol Sci Technol. 1999; 31(6):409-21. [23] Maiman TH. Stimulated optical emis sion in ruby. J Opt Soc Am. 1960; 50(11):113434. [24] Maiman TH. Optical and mi crowave-optical experiments in ruby. Phys Rev Lett. 1960; 4(11):564-66. [25] Maiman TH, Evtuhov V, Hoskins RH, Dh aenens IJ, Asawa CK. Stimulated optical emission in fluorescent solids, spectros copy and stimulated emission in ruby. Physical Review. 1961; 123(4):1151-57. [26] Brech F, Cross L. Optical microemission stimulated by a ruby la ser. Appl Spectrosc. 1962; 16:59. [27] Debras-Gudon J, Liodec N. De l'utilis ation du faisceau d'un amplificateur a ondes lumineuses par mission induite de rayonnement (laser rubis), comme source nergtique pour l'excitation des spectres dmission des lments. CR Acad Sci. 1963; 257:3336-39. [28] Maker PD, Terhune RW, Savage CM. Op tical third harmonic generation. Quantum Electronics. 1964; 2. [29] Runge EF, Minck RW, Bryan FR. Spectrochemical analysis using a pulsed laser source. Spectrochim Acta, Part B.1964; 20:733-35. 145

PAGE 146

[30] Young M, Hercher M, Wu CY. Some char acteristics of laser-induced air sparks. J Appl Phys. 1966; 37(13):4938. [31] Felske A, Hagenah WD, Laqua K. ber ei nige erfahrungen bei der makrospektralanalyse mit laserlichtquellen: durchschnittsanalyse metallischer proben. Spectrochim Acta, Part B. 1972; 27(1):1-4. [32] Afanasev YV, Krokhin ON. Vaporization of matter exposed to laser emission. Soviet Physics JETP. 1967; 25:639-45. [33] Biberman LM, Norman GE. Continuous spect ra of atomic gases and plasma. Sov Phys Uspekhi. 1967; 10:52-54. [34] Raizer IUP. Laser-induc ed discharge phenomena. New York: Consultants Bureau 1977. [35] Raizer YP. Breakdown and heating of gases under the influence of a laser beam. Sov Phys Uspekhi. 1966; 8:650-73. [36] Zeldovich YB, Raizer YP. Physics of shock waves and high-temperature hydrodynamic phenomena. Mineola: Dover 2002. [37] Buzukov AA, Popov YA, Teslenko VS. Expe rimental study of explosion caused by focusing monopulse laser radiation in water. J Appl Mech Tech Phys. 1969; 10(5): 701-08. [38] Lauterborn W. High-speed photography of laser-induced breakdown in liquids. Appl Phys Lett. 1972; 21:27-29. [39] Teslenko VS. Investigation of photoacous tic and photohydrodynamic parameters of laser breakdown in liquids. Sov J Quantum Electron. 1977; 7:981-84. [40] Belyaev EB, Godlevski AP, Kopytin YD. Laser spectroch emical analysis of aerosols. Sov J Quantum Electron. 1978; 8:1459-63. [41] Edwards AL, J. A. Fleck, Jr. Two-dime nsional modeling of aer osol-induced breakdown in air. J Appl Phys. 1979; 50(6):4307-13. [42] Ivanov YV, Kopytin YD. Selective interacti on of a train of laser pulses with an aerosol medium. J Quantum Electron. 1982; 12:355-57. [43] Lencioni DE. The effect of dust on 10.6mu m laser-induced air breakdown. Appl Phys Lett. 1973; 23(1):12-14. [44] Cerrai E, Trucco R. On the matrix eff ect in laser sampled spectrochemical analysis. Energia Nucleare. 1968; 15:581-85. [45] Marich KW, Carr PW, Treytl WJ, Glick D. Effect of matrix material on laser-induced elemental spectral emission. Anal Chem. 1970; 42:1775-79. 146

PAGE 147

[ 46] Treytl WJ, Saffir AJ, Glick D, Marich KW, J. B. Orenberg. Detection limits in analysis of metals in biological materials by laser mi croprobe optical emission spectrometry. Anal Chem. 1972; 44:1903-04. [47] Fichet P, Mauchien P, Moulin C. Determination of impurities in uranium and plutonium dioxides by laser-induced breakdown spectro scopy. Appl Spectrosc. 1999; 53:1111-17. [48] Sneddon J, Lee YI. Application of la ser-induced breakdown spectrometry in urban health. Microchem J. 2000; 67(1-3):201-05. [49] Stratis DN, Eland KL, Carter JC, Tom linson SJ, Angel SM. Comparison of acoustooptic and liquid crystal tuna ble filters for laser-induced breakdown spectroscopy. Appl Spectrosc. 2001; 55:999-1004. [50] Gonzaga FB, Pasquini C. A new detection system for laser induced breakdown spectroscopy based on an acousto-optical tunable filter coupled to a photomultiplier: application for manganese determination in steel. Spectrochim Acta, Part B. 2008; 63(11):1268-73. [51] Houck MM. Mute witnesse s: trace evidence analysis. San Diego, London: Academic Press 2001. [52] Caddy B. Forensic examination of glass and paint analysis an d interpretation. LondonNew York: Taylor & Francis 2001. [53] Zadora G. Glass analysis for forensic pur poses a comparison of classification methods. J Chemom. 2007; 21(5-6):174-86. [54] Curran JM, Hicks TN, Buckleton JS. Fore nsic interpretation of glass evidence. Boca Raton: CRC Press 2000. [55] Tsach T, Wiesner S, Shor Y. Empirical proof of physical match: systematic research with tensile machine. Forensic Sci Int. 2007; 166(1):77-83. [56] Winchell AN, Emmons RC. Some methods for determining refractive indices American Mineral. 1926; 11:115-19. [57] Almirall JR, Cole MD, Gettinby G, Furton KG. Discrimination of glass sources using elemental composition and refractive index: development of predictive models. Science & Justice. 1998; 38(2):93-100. [58] Bennett RL, Kim ND, Curran JM, Coulson SA, Newton AWN. Spatial variation of refractive index in a pane of float gla ss. Science & Justice. 2003; 43(2):71-76. [59] Koons RD, Buscaglia J. Distribution of refractive index values in sheet glasses. Forensic Science Communications. 2001; 3(1). 147

PAGE 148

[60] Koons RD, Buscaglia JA. The forensic significance of glass composition and refractive index measurements. J Forensic Sci. 1999; 44(3):496-503. [61] Koons RD, Peters CA, Rebbert PS. Comparison of refractive-inde x, energy dispersivex-ray fluorescence and inductively coupled plasma atomic emission-spectrometry for forensic characterization of sheet glass fr agments. J Anal At Spectrom. 1991; 6(6):45156. [62] Lambert JA, Evett IW. The refractive-i ndex distribution of control glass samples examined by the forensic-science laboratorie s in the United Kingdom. Forensic Sci Int. 1984; 26(1):1-23. [63] Newton AWN, Buckleton JS. An investigati on into the relationship between edge counts and the variability of the refractive index of glass. Forens ic Sci Int. 2008; 177(1):24-31. [64] Newton AWN, Curran JM, Triggs CM, Buck leton JS. The consequences of potentially differing distributions of the refractive i ndices of glass fragments from control and recovered sources. Forensic Sci Int. 2004; 140(2-3):185-93. [65] Pawluk-Kolc M, Zigba-Palus J, Parczew ski A. The effect of re-annealing on the distribution of refractive i ndex in a windscreen and a windowpane classification of glass samples. Forensic Sci Int. 2008; 174(2-3):222-28. [66] Stoney DA, Thornton JI. The forensic signi ficance of the correlation of density and refractive-index in gla ss evidence. Forensic Sci Int. 1985; 29(3-4):147-57. [67] Suzuki Y, Kasamatsu M, Suzuki S, Naka nishi T, Takatsu M, Muratsu S, Shimoda O, Watanabe S, Nishiwaki Y, Miyamoto N. Forensic discrimination of sheet glass by a refractive-index measurement and elemental analysis with synchr otron radiation x-ray fluorescence spectrometry. An al Sci. 2005; 21(7):855-59. [68] Suzuki Y, Sugita R, Suzuki S, Marumo Y. Forensic discrimination of bottle glass by refractive index measurement and analysis of trace elements with ICP-MS. Anal Sci. 2000; 16(11):1195-98. [69] Buscaglia JA. Elemental analysis of small gl ass fragments in forensic science. Anal Chim Acta. 1994; 288(1-2):17-24. [70] Zurhaar A, Mullings L. Characterization of forensic glass samples using inductively coupled plasma mass-spectrometry. J Anal At Spectrom. 1990; 5(7):611-17. [71] Hughes JC, Catterick T, Southeard G. Th e quantitative analysis of glass by atomic absorption spectroscopy. Forensic Science. 1976; 3:217-27. [72] Blacklock EC, Rogers A, Wall C, Wheal s BB. The quantitative analysis of glass by emission spectrography: a six element survey. Forensic Science. 1976; 2:121-30. 148

PAGE 149

[73] Ducreux-Zappa M, Mermet JM. Analys is of glass by UV laser ablation inductively coupled plasma atomic emission spectrometr y. Effects of the laser parameters on the amount of ablated material and the temporal behavior of the signal for different types of laser. Spectrochim Acta, Part B. 1996; 51(3):321-32. [74] Ducreux-Zappa M, Mermet JM. Analys is of glass by UV laser ablation inductively coupled plasma atomic emission spectrometry. Analytical figures of merit. Spectrochim Acta, Part B. 1996; 51(3):333-41. [75] Hickman DA. A classifica tion scheme for glass. Fore nsic Sci Int. 1981; 17:265-81. [76] Koons RD, Fiedler C, Rawalt RC. Cla ssification and discrimination of sheet and container glasses by inductively couple d plasma-atomic emission-spectrometry and pattern-recognition. J Forens ic Sci. 1988; 33(1):49-67. [77] Ohsaki H, Kasamatsu M, Suzuki Y, Suzuki S. Forensic discrimination of steel wire by determination of trace elements with IC P-AES. Bunseki Kagaku. 2007; 56(12):1191-95. [78] Suzuki Y, Kasamatsu M, Suzuki S, Marumo Y. Forensic discrimination of lead-tin solder based on the trace impurity analysis by ICP-AES. Anal Sci. 2003; 19(3):415-18. [79] Suzuki Y, Kasamatsu M, Suzuki S, Ohas hi K, Kawase A. Forensic discrimination of arsenous acid by the analysis of trace im purities with ICP-AES. Bunseki Kagaku. 2001; 50(5):335-39. [80] Wolnik KL, Gaston CM, Fricke FL. Analysis of glass in product tampering investigations by inductively coupled plasma atomic em ission-spectrometry with a hydrofluoric-acid resistant torch. J Anal At Spectrom. 1989; 4(1):27-31. [81] Verita M, Basso R, Wypyski MT, Koestler RJ. X-ray microanalysis of ancient glassy materials: a comparative study of wave length dispersive a nd energy dispersive techniques. Archaeometr y. 1994; 36(2):241-51. [82] Falcone R, Sommariva G, Verita M. WDXRF, EPMA and SEM/EDX quantitative chemical analyses of small glass samp les. Microchim Acta. 2006; 155(1-2):137-40. [83] Falcone R, Renier A, Verit M. Wavele ngthdispersive xray fl uorescence analysis of ancient glasses. Archaeo metry. 2002; 44(4):531-42. [84] Furukawa K, Fukushima S. Discrimination of beer bottle glass fragments by wavelength dispersive elect ron probe microanalyzer. Repor ts of National Research Institute of Police Science Research on Forensic Science. 1999; 52:89-98. [85] Falcone R, Galinetto P, Messiga B, Negri E, Riccardi MP, Sommariva G, Verita M. Combined SEM-EDX and mu-raman spectroscop y for the characterisation of glass/Alrich refractory interfaces. Microchim Acta. 2008; 161(3-4):381-86. 149

PAGE 150

[86] Andrasko J, Maehly AC. The discrimi nation between samples of window glass by combining physical and chemical techniques. J Forensic Sci. 1978; 22:250-62. [87] Dudley RJ, Howden CR, Taylor TJ, Sma lldon KW. The discrimination and classification of small fragments of window and non-window glasses using ener gy-dispersive x-ray fluorescence spectrometry. X-Ray Spectrom. 1980; 9(3):119-22. [88] Howden CR, Dudley RJ, Smalldon KW. Th e analysis of small glass fragments using energy dispersive x-ray fluorescence spectrometry. J Forensic Sci Soc. 1978; 18(12):99-112. [89] Reeve V, Mathiesen J, Fong W. Elemen tal analysis by energy dispersive X-ray: A significant factor in the forensic analysis of glass. J Forensic Sci. 1976; 21:291-306. [90] Nakanishi T, Nishiwaki Y, Miyamoto N, Shimoda O, Watanabe S, Muratsu S, Takatsu M, Terada Y, Suzuki Y, Kasamatsu M, Suzuki S. Lower limits of detection of synchrotron radiation high-ene rgy x-ray fluorescence spectrom etry and its possibility for the forensic application for discrimination of glass fragments. Forensic Sci Int. 2008; 175(2-3):227-34. [91] Nishiwaki Y, Nakanishi T, Terada Y, Ninomiya T, Nakai I. Nondestructive discrimination of small glass fragments for forensic examination using high energy synchrotron radiation x-ray fluoresce nce spectrometry. X-Ray Spectrom. 2006; 35(3):195-99. [92] Suzuki Y, Kasamatsu M, Sugita R, Ohta H, Suzuki S, Marumo Y. Forensic discrimination of headlight glass by anal ysis of trace impurities with synchrotron radiation x-ray fluorescence spectrom etry and ICP-MS. Bunseki Kagaku. 2003; 52(6):469-74. [93] Suzuki S, Suzuki Y, Ohta H, Kasamats u M, Nakanishi T. Evaluation of a quantitative method for trace impurities in arsenous acids by synchrotron radiation x-ray fluorescence spectrometry and inductively coupled plasma-atomic emission spectrometry. Anal Sci. 2005; 21(7):775-78. [94] Nishiwaki Y, Shimoyama M, Nakanishi T, Ninomiya T, Nakai I. Application of total reflection x-ray fluorescence spectrometry to small glass fragments. Anal Sci. 2006; 22(10):1297-300. [95] Suzuki Y, Kikkawa HS, Kasamatsu M, Higashikawa Y, Suzuki S. Forensic discrimination of sheet glass exposed to high temperature by the determination of trace impurities using ICP-MS. Anal Sci. 2008; 24(6):745-49. [96] Umpierrez S, Trejos T, Neubauer K, Al mirall J. Determination of iron in glass by solution and laser abla tion DRC-ICP-MS. At Spectrosc. 2006; 27(3):76-79. [97] Suzuki Y, Sugita R, Suzuki S, Marumo Y. Determination of trace impurities in windshield glass by ICP-MS. B unseki Kagaku. 1997; 46(10):825-30. 150

PAGE 151

[98] Duckworth DC, Morton SJ, Bayne CK, Koons RD, Montero S, Almirall JR. Forensic glass analysis by ICP-MS: a multi-element assessment of discriminating power via analysis of variance and pairwise comparisons. J Anal At Spectrom. 2002; 17(7):662-68. [99] Montero S, Hobbs AL, French TA, Almira ll AR. Elemental analysis of glass fragments by ICP-MS as evidence of a ssociation: Analysis of a case. J Forensic Sci. 2003; 48(5):1101-07. [100] Neufeld LM. Application of laser ablation ICP-MS to the analysis of forensic glass samples. Spectroscopy. 2005; 20(7):31-36. [101] Parouchais T, Warner IM, Palmer LT, K obus H. The analysis of small glass fragments using inductively coupled plas ma mass spectrometry. J Forens ic Sci. 1996; 41(3):351-60. [102] Duckworth DC, Bayne CK, Morton SJ, Almirall J. Analysis of variance in forensic glass analysis by ICP-MS: variance within the method. J Anal At Spectrom. 2000; 15(7):82128. [103] Ghazi AM, Millette JR. Environmental forensic application of lead isotope ratio determination: a case study using lase r ablation sector ICP-MS. Environmental Forensics. 2004; 5(2):97-108. [104] Aeschliman DB, Bajic SJ, Baldwin DP, H ouk RS. Multivariate patte rn matching of trace elements in solids by laser ablation induc tively coupled plasma-mass spectrometry: Source attribution and preliminary dia gnosis of fractionation. Anal Chem. 2004; 76(11):3119-25. [105] Bajic SJ, Aeschliman DB, Saetveit NJ Baldwin DP, Houk RS. Analysis of glass fragments by laser ablation-inductively coupl ed plasma-mass spectrometry and principal component analysis. J Fore nsic Sci. 2005; 50(5):1123-27. [106] Becker JS, Tenzler D. Studies of LA-ICPMS on quartz glasses at different wavelengths of a Nd:YAG laser. Fresenius J Anal Chem. 2001; 370(5):637-40. [107] Berends-Montero S, Wiarda W, de Joode P, van der Peijl G. Forensic analysis of float glass using laser ablation inductively coupled plasma ma ss spectrometry (LA-ICP-MS): validation of a method. J Anal At Spectrom. 2006; 21(11):1185-93. [108] Bridge CM, Powell J, Steele KL, Williams M, MacInnis JM, Sigman ME. Characterization of automobile float gl ass with laser-induced breakdown spectroscopy and laser ablation inductively coupled plasma mass spectrometry. Appl Spectrosc. 2006; 60(10):1181-87. [109] Hobbs AL, Almirall JR. Trace elemental analysis of automoti ve paints by laser ablationinductively coupled plasma-mass spectrometr y (LA-ICP-MS). Anal Bioanal Chem. 2003; 376(8):1265-71. 151

PAGE 152

[110] Ito M, Hokura A, Oishi M, Nishiwaki Y, Nakai I. Characterization of sheet glass based on trace elements analysis using LA-ICP-MS an d their applications to forensic science. Bunseki Kagaku. 2007; 56(12):1115-25. [111] Latkoczy C, Becker S, Ducking M, Gunt her D, Hoogewerff JA, Almirall JR, Buscaglia J, Dobney A, Koons RD, Montero S, van der Peijl GJQ, Stoecklein WRS, Trejos T, Watling JR, Zdanowicz VS. Development and evaluation of a standard method for the quantitative determination of elements in float glass samples by LA-ICP-MS. J Forensic Sci. 2005; 50(6):1327-41. [112] Scadding CJ, Watling RJ, Thomas AG. The potential of using lase r ablation inductively coupled plasma time of fight mass spect rometry (LA-ICP-TOF-MS) in the forensic analysis of micro debris. Talanta. 2005; 67(2):414-24. [113] Trejos T, Almirall JR. Sampling strategies for the analysis of glass fragments by LA-ICPMS Part II: Sample size and sample shape considerations. Talanta. 2005; 67(2):396401. [114] Trejos T, Almirall JR. Sampling strategies for the analysis of glass fragments by LA-ICPMS Part I. Micro-homogeneity study of glass and its application to the interpretation of forensic evidence. Talanta. 2005; 67(2):388-95. [115] Trejos T, Montero S, Almirall JR. Analys is and comparison of glass fragments by laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) and ICP-MS. Anal Bioanal Chem. 2003; 376(8):1255-64. [116] Watling RJ, Lynch BF, Herring D. Use of la ser ablation inductively coupled plasma mass spectrometry for fingerprinting scene of crime evidence. J Anal At Spectrom. 1997; 12(2):195-203. [117] Becker S, Ducking M, Watzke P, Stoecklein W. Laser ablation in forensic glass analysis: The use of matrix matched standards for qua ntitative float glass analysis. Forensic Sci Int. 2003; 136:361-61. [118] Hordijk M, Wiarda W, de Joode P, van der Peijl G, Montero S. LA-ICP-MS method for the elemental profiling of fore nsic glass: A study of a float glass sample set collected in the Netherlands. Forensic Sci Int. 2003; 136:358-59. [119] Neufeld L. Application of laser ablation ICP-MS to the analysis of forensic glass samples. Spectroscopy. 2005; 20(7). [120] Russo RE, Mao XL, Chan WT, Bryant MF Kinard WF. Laser ablation sampling with inductively coupled plasma atomic emission spectrometry for the analysis of prototypical glasses. J Anal At Sp ectrom. 1995; 10(4):295-301. [121] Dockery CR, Goode SR. Laser-induced breakdown spectroscopy for the detection of gunshot residues on the hands of a s hooter. Appl Opt. 2003; 42(30):6153-58. 152

PAGE 153

[122] Corsi M, Cristoforetti G, Hidalgo M, Le gnaioli S, Palleschi V, Salvetti A, Tognoni E, Vallebona C. Application of laser-induced breakdown spect roscopy technique to hair tissue mineral analysis. Applie d Optics. 2003; 42(30):6133-37. [123] Al-Jeffery MO, Telle HH. LIBS and LIFS for rapid detection of Rb traces in blood. In: Robert RA, editor.; 2002: SPIE; 2002. p. 152-61. [124] Taschuk MT, Tsui YY, Fedosejevs R. Detec tion and mapping of latent fingerprints by laser-induced breakdown spectroscopy. Appl Spectrosc. 2006; 60:1322-27. [125] Page D. Read our LIBS: crime scen e wood analysis gets new bark. Forensic Magazine. 2005; 21(5):15-17. [126] Martin MZ, Labbe N, Andre N, Harris R, Ebinger M, Wullschleger SD, Vass AA. High resolution applications of laser-induced breakdown spectroscopy for environmental and forensic applications. Spectrochim Acta, Part B. 2007; 62(12):1426-32. [127] Almirall J, Umpierrez S, Castro W, Gor nushkin I, Winefordner J. Forensic elemental analysis of materials by laser induced breakdown spectroscopy (LIBS). In: Edward MC, editor.; 2005: SPIE; 2005. p. 657-66. [128] Bridge CM, Powell J, Steele KL, Sigman ME. Forensic comparative glass analysis by laser-induced breakdown spect roscopy. Spectrochim Acta, Part B. 2007; 62(12):1419-25. [129] Barnett C, Cahoon E, Almirall JR. Wavele ngth dependence on the el emental analysis of glass by laser induced breakdown spectro scopy. Spectrochim Acta, Part B. 2008; 63(10):1016-23. [130] Naes BE, Umpierrez S, Ryland S, Barnett C, Almirall JR. A comparison of laser ablation inductively coupled plasma mass spectrome try, micro x-ray fluorescence spectroscopy, and laser induced breakdown spectroscopy fo r the discrimination of automotive glass. Spectrochim Acta Part B. 2008; 63(10):1145-50. [131] Carmona N, Oujja M, Gaspard S, Garcia -Heras M, Villegas MA, Castillejo M. Lead determination in glasses by laser-induced breakdown spectroscopy. Spectrochim Acta, Part B. 2007; 62(2):94-100. [132] Carmona N, Oujja M, Rebollar E, Romich H, Castillejo M. Analysis of corroded glasses by laser induced breakdown spectroscopy. Spectrochim Acta, Part B. 2005; 60(78):1155-62. [133] Klein S, Hildenhagen J, Dickmann K, Stratoudaki T, Za firopulos V. LIBS-spectroscopy for monitoring and control of the laser clean ing process of stone and medieval glass. Journal of Cultural Heritage. 2000; 1(Supplement 1):S287-S92. [134] Klein S, Stratoudaki T, Zafiropulos V, Hildenhagen J, Dickmann K, Lehmkuhl T. Laserinduced breakdown spectroscopy for on-line co ntrol of laser cleani ng of sandstone and stained glass. Appl Phys A: Ma ter Sci Process. 1999; 69(4):441. 153

PAGE 154

[135] Lal BS, Yueh FY, Singh JP. Glass-batch composition monitoring by laser-induced breakdown spectroscopy. Appl Opt. 2005; 44(18):3668-74. [136] Lazic V, Fantoni R, Colao F, Santagata A, Morone A, Spizzichino V. Quantitative laser induced breakdown spectroscopy analysis of ancient marbles and corrections for the variability of plasma parameters and of ablation rate. J Anal At Spectrom. 2004; 19(4):429-36. [137] Muller K, Stege H. Evaluation of the an alytical potential of laser-induced breakdown spectrometry (LIBS) for the analysis of historical glasses. Archaeometry. 2003; 45:42133. [138] Panne U, Clara M, Haisch C, Niessner R. Analysis of glass and glass melts during the vitrification of fly and bottom ashes by la ser-induced plasma spectroscopy. Part II. Process analysis. Spectrochim Ac ta, Part B. 1998; 53(14):1969-81. [139] Su CF, Feng SM, Singh JP, Yuen FY, Rigsby JT, Monts DL, Cook RL. Glass composition measurement using laser induced breakdown spectrometry. Glass Technology. 2000; 41(1):16-21. [140] Yoon Y, Kim T, Yang M, Lee K, Lee G. Qu antitative analysis of pottery glaze by laser induced breakdown spectroscopy. Mi crochem J. 2001; 68(2-3):251-56. [141] Burakov V, Tarasenko N, Nedelko M, Isakov S. Time-resolved spectroscopy and imaging diagnostics of single pulse and coll inear double pulse laser induced plasma from a glass sample. Spectrochim Ac ta, Part B. 2008; 63(1):19-26. [142] Carmona N, Oujja M, Gaspard S, Garcia -Heras M, Villegas MA, Castillejo M. Lead determination in glasses by laser-induced breakdown spectroscopy. Spectrochim Acta Part B. 2007; 62(2):94-100. [143] Carmona N, Oujja M, Rebollar E, Romich H, Castillejo M. Analysis of corroded glasses by laser induced breakdown spectroscopy. Sp ectrochim Acta Part B. 2005; 60(7-8):115562. [144] Craparo JC, Weisberg A, De Saro R. Ne w technique developed to measure glass batch, sort cullet. Am Ceram Soc Bull. 2006; 85(3):33-37. [145] Galiova M, Mozna V, Stankova A, N ovotny K, Kanicky V. Study of laser-sample interactions of glasses an d soils using laser-induced breakdown spectroscopy and optical emission spectroscopy with inductively c oupled plasma. Chem Listy. 2006; 100(3):20409. [146] Gautier C, Fichet P, Menut D, Dubessy J. Applications of the double-pulse la ser-induced breakdown spectroscopy (LIBS) in the collinear beam geometry to the elemental analysis of different materials. Spectroch im Acta, Part B. 2006; 61(2):210-19. 154

PAGE 155

[147] Giakoumaki A, Melessanaki K, Angl os D. Laser-induced breakdown spectroscopy (LIBS) in archaeological scie nce-applications and prospect s. Anal Bioanal Chem. 2007; 387(3):749-60. [148] Klein S, Stratoudaki T, Zafiropulos V, Hildenhagen J, Dickmann K, Lehmkuhl T. Laserinduced breakdown spectroscopy for on-line co ntrol of laser cleani ng of sandstone and stained glass. Appl Phys A: Mater Sci Process. 1999; 69(4):441-44. [149] Lal B, Zheng HB, Yueh FY, Singh JP. Parame tric study of pellets for elemental analysis with laser-induced breakdown spectro scopy. Appl Opt. 2004; 43(13):2792-97. [150] Laville S, Sabsabi M, Doucet FR. Multi-e lemental analysis of solidified mineral melt samples by laser-induced breakdown spectro scopy coupled with a linear multivariate calibration. Spectrochim Acta Part B. 2007; 62(12):1557-66. [151] Loebe K, Uhl A, Lucht H. Microanalysis of tool steel and glass with laser-induced breakdown spectroscopy. Appl Opt. 2003; 42(30):6166-73. [152] Stratis DN, Eland KL, Angel SM. Enhancem ent of aluminum, titanium, and iron in glass using pre-ablation spark dual-pulse LIBS. Appl Spectrosc. 2000; 54(12):1719-26. [153] Tornari V, Zafiropulos V, Bonarou A, Vainos NA, Fotakis C. Modern technology in artwork conservation: a lase r-based approach for process control and evaluation. Optics and Lasers in Engineer ing. 2000; 34(4-6):309-26. [154] Yun JI, Klenze R, Kim JI. Laser-indu ced breakdown spectroscopy for the on-line multielement analysis of highly radioactive glass melt. Part 1: characterization and evaluation of the method. Appl Spectrosc. 2002; 56(4):437-48. [155] Gornushkin IB, Smith BW, Nasajpour H, Winefordner JD. Identification of solid materials by correlation analys is using a microscopic laser-induced plasma spectrometer. Anal Chem. 1999; 71(22):5157-64. [156] Anzano JM, Gornushkin IB, Smith BW Winefordner JD. Laser-induced plasma spectroscopy for plastic identificatio n. Polym Eng Sci. 2000; 40(11):2423-29. [157] Anzano JM, Villoria MA, Gornushkin IB, Smith BW, Winefordner JD. Laser-induced plasma spectroscopy for characterization of archaeological material. Can J Anal Sci Spectros. 2002; 47(5):134-40. [158] Gornushkin IB, Ruiz-Medina A, Anzano JM, Smith BW, Winefordner JD. Identification of particulate materials by correlation analysis using a micr oscopic laser induced breakdown spectrometer. J Anal At Spectrom. 2000; 15(6):581-86. [159] Edwards L. Laser-induced breakdown sp ectroscopy for the determination of carbon in soil. Gainesville: Univ ersity of Florida; 2007. 155

PAGE 156

[160] Yalcin S, Crosley DR, Smith GP, Faris GW Influence of ambient conditions on the laser air spark. Appl Phys B: Lasers Opt. 1999; 68(1):121-30. [161] Wachter JR, Cremers DA. Determination of uranium in solution using laser-induced breakdown spectroscopy. Appl Spectrosc. 1987; 41:1042-48. [162] Multari RA, Foster LE, Cremers DA, Ferris MJ. Effect of sampling geometry on elemental emissions in laser-induced br eakdown spectroscopy. Appl Spectrosc. 1996; 50(12):1483-99. [163] Kempenaers L, Janssens K, Jochum KP, Vincze L, Vekemans B, Somogyi A, Drakopoulos M, Adams F. Mi cro-heterogeneity study of tr ace elements in USGS, MPIDING and NIST glass reference materials by means of synchrotron micro-XRF. J Anal At Spectrom. 2003; 18(4):350-57. [164] Ott RL, Longnecker MT. A first course in statistical methods. Belmont: Duxbury Press 2003. [165] Galbcs G, Gornushkin IB, Smith BW, Wine fordner JD. Semi-quantitative analysis of binary alloys using laser-induced breakdown spectroscopy and a new calibration approach based on linear co rrelation. Spectrochim Acta, Part B. 2001; 56(7):1159-73. [166] Galbcs G, Gornushkin IB, Winefordner JD. Generalization of a new calibration method based on linear correlation. Talanta. 2004; 63(2):351-57. [167] Gornushkin IB, Mueller M, Panne U, Wine fordner JD. Insights into linear and rank correlation for material identification in laser-induced breakdown spectroscopy and other spectral techniques. Appl Spectrosc. 2008; 62:542-53. [168] Personal communication with Dr. Jose Almirall. March, 2008. [169] De Lucia Jr FC, Gottfried JL, Munson CA, Miziolek AW. Double pulse laser-induced breakdown spectroscopy of explosives: Initia l study towards improved discrimination. Spectrochim Acta, Part B. 2007; 62(12):1399-404. [170] DeLucia FC, Samuels AC, Harmon RS, Wa lters RA, McNesby KL, LaPointe A, Winkel RJ, Miziolek AW. Laser-induced breakdown spectroscopy (LIBS): A promising versatile chemical sensor technology for hazardous material detection. IEEE Sens J. 2005; 5(4):681-89. [171] Hybl JD, Lithgow GA, Buckley SG. La ser-induced breakdown spectroscopy detection and classification of biol ogical aerosols. Appl Spect rosc. 2003; 57(10):1207-15. [172] Munson CA, De Lucia JFC, Piehler T, McNesby KL, Miziolek AW. Investigation of statistics strategies for improving the disc riminating power of laser-induced breakdown spectroscopy for chemical and biological warfare agent simulants. Spectrochim Acta, Part B. 2005; 60(7-8):1217-24. 156

PAGE 157

[173] Samuels AC, DeLucia FC, McNesby KL, Miziolek AW. Laser-induced breakdown spectroscopy of bacterial spores, molds, pollens, and protein: initial studies of discrimination potential. A ppl Opt. 2003; 42(30):6205-09. [174] Jurado-Lopez A, De Castro MDL. Ra nk correlation of la ser-induced breakdown spectroscopic data for the identification of alloys used in jewelry manufacture. Spectrochim Acta, Part B. 2003; 58(7):1291-99. [175] Lopez AJ, Nicolas G, Mateo MP, Ramil A, Pinon V, Yanez A. LIPS and linear correlation analysis applied to the classifi cation of roman pottery terra sigillata. Appl Phys A: Mater Sci Process. 2006; 83(4):695-98. [176] Sattmann R, Monch I, Krause H, Noll R, Couris S, Hatziapostolou A, Mavromanolakis A, Fotakis C, Larrauri E, Miguel R. Lase r-induced breakdown spectroscopy for polymer identification. Appl Spectrosc. 1998; 52(3):456-61. [177] Rodriguez-Celis EM, Gornushkin IB, Heitmann UM, Almirall JR, Smith BW, Winefordner JD, Omenetto N. Laser induced breakdown spectroscopy as a tool for discrimination of glass for forensic applic ations. Anal Bioanal Chem. 2008; 391(5):196168. [178] Ismail MA, Cristoforetti G, Legnaioli S, Pa rdini L, Palleschi V, Salvetti A, Tognoni E, Harith MA. Comparison of detection limits, for two metallic matrices, of laser-induced breakdown spectroscopy in the single and doubl e-pulse configurations. Anal Bioanal Chem. 2006; 385(2):316-25. [179] Ismail MA, Imam H, Elhassan A, Youniss WT, Harith MA. LIBS limit of detection and plasma parameters of some elements in two different metallic matrices. J Anal At Spectrom. 2004; 19(4):489-94. [180] Sturm V, Peter L, Noll R. Steel analys is with laser-induced breakdown spectrometry in the vacuum ultraviolet. Appl Spectrosc. 2000; 54(9):1275-78. [181] Sturm V, Vrenegor J, Noll R, Hemmerlin M. Bulk analysis of steel samples with surface scale layers by enhanced laser ablation and LIBS analysis of C, P, S, Al, Cr, Cu, Mn and Mo. J Anal At Spectrom. 2004; 19(4):451-56. [182] Panne U, Haisch C, Clara M, Niessner R. Analysis of glass and glass melts during the vitrification process of fl y and bottom ashes by laser-induced plasma spectroscopy. Part I: normalization and plasma diagnostics. Spectrochim Acta, Part B. 1998; 53(14):195768. [183] Le Drogoff B, Margot J, Chaker M, Sabs abi M, Barthlemy O, Johnston TW, Laville S, Vidal F, von Kaenel Y. Temporal characte rization of femtosecond laser pulses induced plasma for spectrochemical analysis of alum inum alloys. Spectrochim Acta, Part B. 2001; 56(6):987-1002. 157

PAGE 158

[184] Aguilera JA, Bengoechea J, Aragon C. Spa tial characterization of laser induced plasmas obtained in air and argon with different laser focusing distances. Spectrochim Acta, Part B. 2004; 59(4):461-69. [185] Davison S, Newton RG. Conservation and restoration of glass. Oxford: ButterworthHeinemann 2003. [186] Teixeira S, Bernardin AM. Development of TiO2 white glazes for ceramic tiles. Dyes and Pigments. 2009; 80(3):292-96. [187] Bray C. Dictionary of glass: material s and techniques. Tortol a: Craftsman House 1995. [188] Peterson S. The craft and art of clay. London: L. King Pub. 1999. [189] Bou E, Moreno A, Escardino A, Gozalbo A. Microstructural study of opaque glazes obtained from frits of the system: SiO2-Al2O3-B2O3-(P2O5)-CaO-K2O-TiO2. J Eur Ceram Soc. 2007; 27(2-3):1791-96. [190] Herrera KK. From sample to signal in laser-induced breakdown spectroscopy an experimental assessment of existing algor ithms and theoretical modeling approaches. Gainesville, Fla.: University of Florida; 2008. [191] Hans Netten, Young IT, Vliet LJv, Tanke HJ, Willem HV, Sloos CR. Fish and chips: automation of fluorescent dot counting in interphase cell nuclei. Cytometry. 1997; 28(1):1-10. [192] Zhimin Z, Pain B, Fossum ER. Frame-tr ansfer CMOS active pixel sensor with pixel binning. Electron Devices, IEEE Tr ansactions on. 1997; 44(10):1764-68. [193] Sabsabi M, Cielo P. Quantitative anal ysis of aluminum alloys by laser-induced breakdown spectroscopy and plasma characteri zation. Appl Spectrosc. 1995; 49(4):499507. [194] Mohamed WTY. Improved LIBS limit of de tection of Be, Mg, Si, Mn, Fe and Cu in aluminum alloy samples using a portable ech elle spectrometer with ICCD camera. Optics & Laser Technology. 2008; 40(1):30-38. [195] Ralchenko Y, Kramida AE, Reader J. NIST atomic spectra database (version 3.1.5). 2008 [cited 2009 February 27]; Available from: http://physics.nist.gov/asd3 [196] lvarez-Trujillo LA, Ferrero A, Laserna JJ, Hahn DW. Alternativ e statistical methods for spectral data processing: applications to laser-induced breakdown spectroscopy of gaseous and aerosol systems. A ppl Spectrosc. 2008; 62(10):1144-52. [197] Gondal MA, Hussain T. Determination of poi sonous metals in wastewater collected from paint manufacturing plant using laser-induc ed breakdown spectroscopy. Talanta. 2007; 71(1):73-80. 158

PAGE 159

[198] Koskelo A, Cremers DA. RCRA materi als analysis by laser-induced breakdown spectroscopy: Detection limits in soils. 1994:1-64. [199] Yamamoto KY, Cremers DA, Ferris MJ, Foster LE. Detection of metals in the environment using a portable laser-indu ced breakdown spectroscopy instrument. Appl Spectrosc. 1996; 50:222-33. [200] Cho H-H, Kim Y-J, Jo Y-S, Kitagawa K, Arai N, Lee Y-I. Applic ation of laser-induced breakdown spectrometry for direct determinati on of trace elements in starch-based flours. J Anal At Spectrom. 2001; 16(6):622-27. [201] Kurniawan H, Nakajima S, Batubara JE, Marpaung M, Okamoto M, Kagawa K. Laserinduced shock wave plasma in glass and its application to elemental analysis. Appl Spectrosc. 1995; 49(8):1067-72. [202] Gautier C, Fichet P, Menut D, Lacour J-L, L'Hermite D, Dubessy J. Study of the doublepulse setup with an or thogonal beam geometry for laser-induced breakdown spectroscopy. Spectrochim Acta Part B. 2004; 59(7):975-86. [203] Gornushkin SI, Gornushkin IB, Anzano JM, Smith BW, Winefordner JD. Effective normalization technique for correction of matrix effect s in laser-induced breakdown spectroscopy detection of magnesium in powdered samples. Appl Spectrosc. 2002; 56:433-36. [204] Eppler AS, Cremers DA, Hickmott DD, Fe rris MJ, Koskelo AC. Matrix effects in the detection of Pb and Ba in soils using laser-induced breakdown spectroscopy. Appl Spectrosc. 1996; 50(9):1175-81. [205] Chaleard C, Mauchien P, Andre N, Uebbi ng J, Lacour JL, Geertsen C. Correction of matrix effects in quantitative elemental analysis with laser ablation optical emission spectrometry. J Anal At Spectrom. 1997; 12(2):183-88. [206] Ko JB, Sdorra W, Niemax K. On the internal standardizatio n in optical-emission spectrometry of microplasmas produced by la ser ablation of solid samples. Fresenius Zeitschrift Fur Analytisch e Chemie. 1989; 335(7):648-51. [207] Ciucci A, Corsi M, Palleschi V, Rastel li S, Salvetti A, Tognoni E. New procedure for quantitative elemental analysis by laser-i nduced plasma spectroscopy. Appl Spectrosc. 1999; 53(8):960-64. [208] Mohamed WTY. Calibration free lase r-induced breakdown spectroscopy (LIBS) identification of seawater salinity. Optica Applicata. 2007; 37(1-2):5-19. [209] Herrera KK, Tognoni E, Omenetto N, Sm ith BW, Winefordner JD. Semi-quantitative analysis of metal alloys, brass and soil samples by calibration-free laser-induced breakdown spectroscopy: recent results and c onsiderations. J Anal At Spectrom. 2009:-. 159

PAGE 160

[210] Bulajic D, Corsi M, Cristoforetti G, Legnaioli S, Palleschi V, Salvetti A, Tognoni E. A procedure for correcting self-absorption in calibration free-las er induced breakdown spectroscopy. Spectrochim Acta, Part B. 2002; 57(2):339-53. [211] Ciucci A, Palleschi V, Rastelli S, Salv etti A, Singh DP, Tognoni E. CF-LIPS: A new approach to LIPS spectra analysis. Lase r and Particle Beams. 1999; 17(4):793-97. [212] Corsi M, Cristoforetti G, Hidalgo M, Le gnaioli S, Palleschi V, Salvetti A, Tognoni E, Vallebona C. Double pulse, calibration-free laser-induced breakdown spectroscopy: A new technique for in situ standard-less analysis of pollute d soils. Appl Geochem. 2006; 21(5):748-55. [213] Pandhija S, Rai AK. In situ multielement al monitoring in coral skeleton by CF-LIBS. Appl Phys B: Lasers Opt. 2009; 94(3):545-52. [214] Tognoni E, Cristoforetti G, Legnaloli S, Palleschi V, Salvetti A, Mueller M, Panne U, Gomushkin I. A numerical study of expected accuracy and precision in Calibration-Free Laser-Induced Breakdown Spectroscopy in th e assumption of ideal analytical plasma. Spectrochim Acta, Part B. 2007; 62(12):1287-302. [215] Yaroshchyk P, Body D, Morrison RJS, Ch adwick BL. A semi-quantitative standard-less analysis method for laser-induced breakdow n spectroscopy. Spectrochim Acta, Part B. 2006; 61(2):200-09. [216] Wang L, Zhang C, Feng Y. Controlled ca libration method for laser induced breakdown spectroscopy. Chin Opt Lett. 2008; 6(1):5-8. [217] Pandhija S, Rai AK. In situ multielement al monitoring in coral skeleton by CF-LIBS. Appl Phys B: Lasers Opt. 2009; 94(3):545-52. [218] Singh VK, Singh V, Rai AK, Thakur SN, Ra i PK, Singh JP. Quantitative analysis of gallstones using laser-induced breakdow n spectroscopy. Appl Opt. 2008; 47(31):G38G47. [219] Burakov VS, Raikov SN. Quantitative analysis of alloys and glasses by a calibration-free method using laser-induced breakdown spectroscopy. Spectrochim Acta, Part B. 2007; 62(3):217-23. [220] Xu L, Bulatov V, Gridin VV, Schechter I. Absolute analysis of pa rticulate materials by laser-induced breakdown spectroscopy. Anal Chem. 1997; 69(11):2103-08. [221] Gornushkin IB, Smith BW, Potts GE, Omene tto N, Winefordner JD. Some considerations on the correlation between signal and background in laser-induced breakdown spectroscopy using single-shot analys is. Anal Chem. 1999; 71(23):5447-49. [222] Carranza JE, Hahn DW. Samp ling statistics and c onsiderations for single-shot analysis using laser-induced breakdown spectroscopy. Spectrochim Acta, Part B. 2002; 57(4):77990. 160

PAGE 161

[223] Gruber J, Heitz J, Arnold N, uerle D, Rama seder N, Meyer W, Hochrtler J, Koch F. In situ analysis of metal melts in metallurgic vacuum devices by laser-induced breakdown spectroscopy. Appl Spectrosc. 2004; 58:457-62. 161

PAGE 162

162 BIOGRAPHICAL SKETCH Esperanza Mariela Rodriguez Celis, better known as Mariela Rodriguez, was born in Peru in 1979 and is the oldest of th ree siblings. In 2002, she gradua ted from Pontifical Catholic University with a bachelor of science in chem istry. In August 2004 she moved to Gainesville, Florida to attend graduate school at University of Florida. She joined the research group of Dr. Nicolo Omenetto in the analytical chem istry program and conducted research while simultaneously taking classes in forensic drug chemistry and forensic toxicology. In December 2007, she graduated with a Master of Science in Pharmacy and Pharmaceutical Sciences with concentration in Forensic Drug Chemistry. Sh e completed her doctoral research in May 2009. Mariela has been married to Oscar Arroyo since 2006. They expected to move back to Canada, shortly after her graduation.